Study guide

4.4 Multivariable Extrema, Hessian Matrix, and Lagrange Multipliers

GRE Subject Test: Mathematics Exam

Track your progress as you read

Finished sections get a check mark.

1 Chapter 1: Single-Variable Differential Calculus

1.1 Limits, Continuity, and Epsilon-Delta Definitions 1.2 Differentiation Rules, Implicit & Inverse Differentiation 1.3 Mean Value Theorem, Rolle's Theorem, and Function Extrema 1.4 L'Hôpital's Rule, Curve Sketching, and Optimization 1.5 Functions, Graphs, Transformations, and Inverses 1.6 Trigonometric Functions, Identities, and Equations 1.7 Coordinate Geometry, Lines, Circles, and Conics 1.8 Parametric and Polar Curves

2 Chapter 2: Single-Variable Integral Calculus

2.1 Riemann Sums and the Fundamental Theorem of Calculus 2.2 Advanced Integration Techniques 2.3 Geometric Applications: Areas, Volumes, and Arc Length 2.4 Improper Integrals and Convergence Criteria

3 Chapter 3: Sequences, Infinite Series & Power Series

3.1 Sequences, Monotonicity, and Limits 3.2 Infinite Series and Convergence Tests 3.3 Power Series, Radius and Interval of Convergence 3.4 Taylor and Maclaurin Series with Error Bounds

4 Chapter 4: Multivariable Differential Calculus

4.1 Multivariable Functions, Limits, and Partial Derivatives 4.2 Directional Derivatives, Gradients, and Tangent Planes 4.3 Multivariable Chain Rule and Implicit Differentiation 4.4 Multivariable Extrema, Hessian Matrix, and Lagrange Multipliers

5 Chapter 5: Multivariable Integration & Vector Calculus

5.1 Multiple Integrals in Cylindrical and Spherical Coordinates 5.2 Change of Variables and the Jacobian Determinant 5.3 Vector Fields, Line Integrals, and Green's Theorem 5.4 Surface Integrals, Stokes' Theorem, and the Divergence Theorem

6 Chapter 6: Ordinary Differential Equations

6.1 First-Order ODEs: Separable, Linear, and Exact Equations 6.2 Special First-Order ODEs: Homogeneous, Bernoulli, and Orthogonal Trajectories 6.3 Second-Order Linear Homogeneous ODEs with Constant Coefficients 6.4 Nonhomogeneous Linear ODEs: Undetermined Coefficients and Variation of Parameters

7 Chapter 7: Elementary Algebra & Number Theory

7.1 Elementary Algebra, Polynomials, Roots, and Factorization 7.2 Divisibility, Primes, and the Euclidean Algorithm 7.3 Modular Arithmetic, Chinese Remainder Theorem, and Euler's Totient

8 Chapter 8: Linear Algebra: Vector Spaces & Transformations

8.1 Vector Spaces, Subspaces, Span, and Linear Independence 8.2 Basis, Dimension, and Coordinate Representations 8.3 Linear Transformations, Kernel, Range, and the Rank-Nullity Theorem 8.4 Matrix Representations of Transformations and Change of Basis

9 Chapter 9: Linear Algebra: Matrices, Determinants & Spectral Theory

9.1 Systems of Linear Equations, Matrix Rank, and Row Reduction 9.2 Determinants, Matrix Inverses, and Cramer's Rule 9.3 Eigenvalues, Eigenvectors, and Diagonalization 9.4 Inner Product Spaces, Gram-Schmidt, and the Spectral Theorem

10 Chapter 10: Abstract Algebra: Groups, Rings & Fields

10.1 Groups, Subgroups, and Permutations 10.2 Cosets, Lagrange's Theorem, Normal Subgroups, and Homomorphisms 10.3 Rings, Ideals, Modules, Polynomial Rings, and Fields

11 Chapter 11: Introductory Real Analysis

11.1 The Real Field and Metric Space Topology 11.2 Compactness, Completeness, and Connectedness in R^n 11.3 Continuity, Uniform Continuity, and Value Theorems 11.4 Sequences and Series of Functions: Pointwise and Uniform Convergence

12 Chapter 12: Complex Analysis

12.1 Complex Numbers and Analytic Functions 12.2 Contour Integration and Cauchy's Theorems 12.3 Laurent Series, Singularities, and the Residue Theorem

13 Chapter 13: Topology, Geometry & Numerical Analysis

13.1 Topological Spaces, Open/Closed Sets, and Bases 13.2 Euclidean Geometry, Similarity, Circles, and Transformations 13.3 Numerical Analysis: Error, Root Finding, Interpolation, and Quadrature

14 Chapter 14: Probability, Statistics & Discrete Mathematics

14.1 Logic, Set Theory, Algorithms, and Combinatorics 14.2 Graph Theory: Trees, Cycles, Paths, and Planarity 14.3 Probability Spaces, Conditional Probability, and Bayes' Theorem 14.4 Random Variables, Expected Value, Variance, and Common Distributions

4.4 Multivariable Extrema, Hessian Matrix, and Lagrange Multipliers

Key Takeaways Explain This Topic AI

Critical points of a differentiable scalar field occur where ∇f(x_0) = 0 or where ∇f fails to exist.

In two dimensions, the Second Derivative Test classifies critical points via discriminant D = f_{xx} f_{yy} - (f_{xy})^2: D > 0 and f_{xx} > 0 yields a local minimum; D > 0 and f_{xx} < 0 yields a local maximum; D < 0 yields a saddle point; D = 0 is inconclusive.

In n dimensions, the nature of a critical point is governed by the eigenvalues of the symmetric Hessian matrix: all positive indicates a local minimum, all negative indicates a local maximum, and mixed signs indicate a saddle point.

Global extrema on a compact set are determined by testing critical points in the interior and optimizing along the boundary via parameterization or Lagrange multipliers (Extreme Value Theorem).

The method of Lagrange multipliers optimizes f(x) subject to k constraints g_1 = c_1, ..., g_k = c_k by solving ∇f = ∑_{i=1}^k λ_i ∇g_i with linearly independent constraint gradients.

Last updated: September 2026

4.4 Multivariable Extrema, Hessian Matrix, and Lagrange Multipliers

Optimization links multivariable calculus, linear algebra, and geometry. On the GRE Mathematics Subject Test, questions test critical point classification via the Hessian matrix, boundary optimization on compact domains, and Lagrange multipliers under geometric constraints.

Critical Points of Multivariable Functions

Let f : D ⊂ R n → R f: D \subset \mathbb{R}^n \to \mathbb{R} f : D ⊂ R n → R . An interior point x 0 ∈ int ⁡ ( D ) \mathbf{x}_0 \in \operatorname{int}(D) x 0 ​ ∈ int ( D ) is a critical point of f f f if:

∇ f ( x 0 ) = 0 \nabla f(\mathbf{x}_0) = \mathbf{0} ∇ f ( x 0 ​ ) = 0 (all partial derivatives vanish simultaneously), or

At least one partial derivative fails to exist at x 0 \mathbf{x}_0 x 0 ​ .

Multivariable Fermat's Theorem

If f f f attains a local extremum at an interior point x 0 \mathbf{x}_0 x 0 ​ and is differentiable there, then ∇ f ( x 0 ) = 0 \nabla f(\mathbf{x}_0) = \mathbf{0} ∇ f ( x 0 ​ ) = 0 . This is a necessary condition , not a sufficient one: a critical point can be a local minimum, local maximum, or saddle point .

The Second Derivative Test in Two Dimensions

Let ( x 0 , y 0 ) (x_0, y_0) ( x 0 ​ , y 0 ​ ) be a critical point of a C 2 C^2 C 2 function f ( x , y ) f(x, y) f ( x , y ) where ∇ f ( x 0 , y 0 ) = 0 \nabla f(x_0, y_0) = \mathbf{0} ∇ f ( x 0 ​ , y 0 ​ ) = 0 . The Hessian matrix is: H ( x , y ) = ( f x x f x y f y x f y y ) H(x, y) = \begin{pmatrix} f_{xx} & f_{xy} \\ f_{yx} & f_{yy} \end{pmatrix} H ( x , y ) = ( f xx ​ f y x ​ ​ f x y ​ f yy ​ ​ )

By Clairaut's Theorem, f x y = f y x f_{xy} = f_{yx} f x y ​ = f y x ​ . The Hessian discriminant is: D = det ⁡ ( H ( x 0 , y 0 ) ) = f x x f y y − ( f x y ) 2 D = \det(H(x_0, y_0)) = f_{xx} f_{yy} - (f_{xy})^2 D = det ( H ( x 0 ​ , y 0 ​ )) = f xx ​ f yy ​ − ( f x y ​ ) 2

Classification Rules

Local Minimum : D > 0 D > 0 D > 0 and f x x > 0 f_{xx} > 0 f xx ​ > 0 (also implies f y y > 0 f_{yy} > 0 f yy ​ > 0 ).

Local Maximum : D > 0 D > 0 D > 0 and f x x < 0 f_{xx} < 0 f xx ​ < 0 (also implies f y y < 0 f_{yy} < 0 f yy ​ < 0 ).

Saddle Point : D < 0 D < 0 D < 0 . The surface curves upwards in some directions and downwards in others.

Inconclusive : D = 0 D = 0 D = 0 . The test provides no information; analyze higher-order terms or directional paths.

The Hessian Matrix and Definiteness in R n \mathbb{R}^n R n

For f : R n → R f: \mathbb{R}^n \to \mathbb{R} f : R n → R , the second-order Taylor expansion about a critical point x 0 \mathbf{x}_0 x 0 ​ is: f ( x 0 + h ) = f ( x 0 ) + 1 2 h T H ( x 0 ) h + R 2 ( h ) f(\mathbf{x}_0 + \mathbf{h}) = f(\mathbf{x}_0) + \frac{1}{2} \mathbf{h}^T H(\mathbf{x}_0) \mathbf{h} + R_2(\mathbf{h}) f ( x 0 ​ + h ) = f ( x 0 ​ ) + 2 1 ​ h T H ( x 0 ​ ) h + R 2 ​ ( h )

Because H ( x 0 ) H(\mathbf{x}_0) H ( x 0 ​ ) is real symmetric, all eigenvalues λ 1 , … , λ n \lambda_1, \dots, \lambda_n λ 1 ​ , … , λ n ​ are real:

Positive Definite (all λ i > 0 \lambda_i > 0 λ i ​ > 0 ): x 0 \mathbf{x}_0 x 0 ​ is a strict local minimum .

Negative Definite (all λ i < 0 \lambda_i < 0 λ i ​ < 0 ): x 0 \mathbf{x}_0 x 0 ​ is a strict local maximum .

Indefinite (mixed signs: λ i > 0 \lambda_i > 0 λ i ​ > 0 and λ j < 0 \lambda_j < 0 λ j ​ < 0 ): x 0 \mathbf{x}_0 x 0 ​ is a saddle point .

Semidefinite (some λ i = 0 \lambda_i = 0 λ i ​ = 0 ): The test is inconclusive .

Sylvester's Criterion

Let Δ k \Delta_k Δ k ​ denote the k k k -th leading principal minor of H H H :

Positive definite ⟺ Δ k > 0 \iff \Delta_k > 0 ⟺ Δ k ​ > 0 for all k = 1 , … , n k = 1, \dots, n k = 1 , … , n .

Negative definite ⟺ ( − 1 ) k Δ k > 0 \iff (-1)^k \Delta_k > 0 ⟺ ( − 1 ) k Δ k ​ > 0 for all k = 1 , … , n k = 1, \dots, n k = 1 , … , n (alternating signs Δ 1 < 0 , Δ 2 > 0 , Δ 3 < 0 , … \Delta_1 < 0, \Delta_2 > 0, \Delta_3 < 0, \dots Δ 1 ​ < 0 , Δ 2 ​ > 0 , Δ 3 ​ < 0 , … ).

Global Extrema on Compact Sets

By the Extreme Value Theorem (EVT) , a continuous function on a compact set K ⊂ R n K \subset \mathbb{R}^n K ⊂ R n attains an absolute maximum and minimum.

Systematic Optimization Procedure

Interior : Identify critical points in int ⁡ ( K ) \operatorname{int}(K) int ( K ) and evaluate f f f .

Boundary : Find extreme values on boundary ∂ K \partial K ∂ K via parameterization or Lagrange multipliers, including corner points.

Comparison : The largest value is the global maximum; the smallest is the global minimum.

  • The Method of Lagrange Multipliers
  • Single Constraint

To optimize f ( x , y , z ) f(x, y, z) f ( x , y , z ) subject to g ( x , y , z ) = c g(x, y, z) = c g ( x , y , z ) = c (with ∇ g ≠ 0 \nabla g \neq \mathbf{0} ∇ g  = 0 ): ∇ f = λ ∇ g , g ( x , y , z ) = c \nabla f = \lambda \nabla g, \quad g(x, y, z) = c ∇ f = λ ∇ g , g ( x , y , z ) = c

Geometric Meaning : At an optimum, the level set of f f f is tangent to g = c g = c g = c . Their normal vectors are collinear, so ∇ f = λ ∇ g \nabla f = \lambda \nabla g ∇ f = λ ∇ g . The scalar λ \lambda λ is the Lagrange multiplier .

Multiple Constraints

To optimize f ( x ) f(\mathbf{x}) f ( x ) subject to g 1 ( x ) = c 1 , … , g k ( x ) = c k g_1(\mathbf{x}) = c_1, \dots, g_k(\mathbf{x}) = c_k g 1 ​ ( x ) = c 1 ​ , … , g k ​ ( x ) = c k ​ : ∇ f ( x ) = ∑ i = 1 k λ i ∇ g i ( x ) \nabla f(\mathbf{x}) = \sum_{i=1}^k \lambda_i \nabla g_i(\mathbf{x}) ∇ f ( x ) = i = 1 ∑ k ​ λ i ​ ∇ g i ​ ( x )

provided constraint gradients ∇ g 1 , … , ∇ g k \nabla g_1, \dots, \nabla g_k ∇ g 1 ​ , … , ∇ g k ​ are linearly independent.

  • Step-by-Step Worked Problems
  • Problem 1: Critical Point Classification

Find and classify all critical points of f ( x , y ) = x 3 + y 3 − 3 x y f(x, y) = x^3 + y^3 - 3xy f ( x , y ) = x 3 + y 3 − 3 x y .

Solution :

Partials: f x = 3 x 2 − 3 y = 0 ⟹ y = x 2 f_x = 3x^2 - 3y = 0 \implies y = x^2 f x ​ = 3 x 2 − 3 y = 0 ⟹ y = x 2 ; f y = 3 y 2 − 3 x = 0 ⟹ x = y 2 f_y = 3y^2 - 3x = 0 \implies x = y^2 f y ​ = 3 y 2 − 3 x = 0 ⟹ x = y 2 .

Substitute: x = x 4 ⟹ x ( x 3 − 1 ) = 0 x = x^4 \implies x(x^3 - 1) = 0 x = x 4 ⟹ x ( x 3 − 1 ) = 0 , giving critical points ( 0 , 0 ) (0, 0) ( 0 , 0 ) and ( 1 , 1 ) (1, 1) ( 1 , 1 ) .

Second partials: f x x = 6 x , f y y = 6 y , f x y = − 3 ⟹ D = 36 x y − 9 f_{xx} = 6x, f_{yy} = 6y, f_{xy} = -3 \implies D = 36xy - 9 f xx ​ = 6 x , f yy ​ = 6 y , f x y ​ = − 3 ⟹ D = 36 x y − 9 .

Classification:

At ( 0 , 0 ) (0, 0) ( 0 , 0 ) : D = − 9 < 0 ⟹ D = -9 < 0 \implies D = − 9 < 0 ⟹ Saddle point with f ( 0 , 0 ) = 0 f(0, 0) = 0 f ( 0 , 0 ) = 0 .

At ( 1 , 1 ) (1, 1) ( 1 , 1 ) : D = 27 > 0 D = 27 > 0 D = 27 > 0 and f x x = 6 > 0 ⟹ f_{xx} = 6 > 0 \implies f xx ​ = 6 > 0 ⟹ Local minimum with f ( 1 , 1 ) = − 1 f(1, 1) = -1 f ( 1 , 1 ) = − 1 .

Problem 2: Lagrange Multiplier Optimization

Find the maximum value of f ( x , y , z ) = x + 2 y + 3 z f(x, y, z) = x + 2y + 3z f ( x , y , z ) = x + 2 y + 3 z on x 2 + y 2 + z 2 = 14 x^2 + y^2 + z^2 = 14 x 2 + y 2 + z 2 = 14 .

Solution :

Gradients: ∇ f = ⟨ 1 , 2 , 3 ⟩ \nabla f = \langle 1, 2, 3 \rangle ∇ f = ⟨ 1 , 2 , 3 ⟩ and ∇ g = ⟨ 2 x , 2 y , 2 z ⟩ \nabla g = \langle 2x, 2y, 2z \rangle ∇ g = ⟨ 2 x , 2 y , 2 z ⟩ .

Set ∇ f = λ ∇ g \nabla f = \lambda \nabla g ∇ f = λ ∇ g : 1 = 2 λ x , 2 = 2 λ y , 3 = 2 λ z ⟹ x = 1 2 λ , y = 1 λ , z = 3 2 λ 1 = 2\lambda x, 2 = 2\lambda y, 3 = 2\lambda z \implies x = \frac{1}{2\lambda}, y = \frac{1}{\lambda}, z = \frac{3}{2\lambda} 1 = 2 λ x , 2 = 2 λ y , 3 = 2 λ z ⟹ x = 2 λ 1 ​ , y = λ 1 ​ , z = 2 λ 3 ​ .

Substitute: 1 + 4 + 9 4 λ 2 = 14 ⟹ 4 λ 2 = 1 ⟹ λ = ± 1 2 \frac{1 + 4 + 9}{4\lambda^2} = 14 \implies 4\lambda^2 = 1 \implies \lambda = \pm \frac{1}{2} 4 λ 2 1 + 4 + 9 ​ = 14 ⟹ 4 λ 2 = 1 ⟹ λ = ± 2 1 ​ .

Candidates: λ = 1 / 2 ⟹ ( 1 , 2 , 3 ) \lambda = 1/2 \implies (1, 2, 3) λ = 1/2 ⟹ ( 1 , 2 , 3 ) with f = 14 f = 14 f = 14 ; λ = − 1 / 2 ⟹ ( − 1 , − 2 , − 3 ) \lambda = -1/2 \implies (-1, -2, -3) λ = − 1/2 ⟹ ( − 1 , − 2 , − 3 ) with f = − 14 f = -14 f = − 14 .

The maximum value is 14 14 14 .

GRE Exam Traps & Pitfalls

Trap 1: Assuming Inconclusive Tests ( D = 0 D = 0 D = 0 ) Imply Saddle Points When D = 0 D = 0 D = 0 , the Second Derivative Test is inconclusive. Do not guess saddle point; check higher-order terms.

Trap 2: Ignoring Boundaries in Global Extrema On compact domains, global extrema frequently occur on the boundary rather than at interior critical points.

Trap 3: Division by Zero in Lagrange Multipliers When solving ∇ f = λ ∇ g \nabla f = \lambda \nabla g ∇ f = λ ∇ g , check if setting a variable to zero provides a solution before dividing by it.

Loading diagram... Multivariable Critical Point Classification and Optimization Flowchart