[ Main | Function | x-Partial | x-Zeros | y-Partial | y-Zeros | Solution | Hessian | Classification ]

Critical Points of a Real-Valued Function of Two Variables
by Tom Davis1


At a critical or stationary point on a surface, the tangent plane is horizontal, analogous to a horizontal tangent line for critical points on a curve.
No support for LM Objects Given a twice continuously differentiable or C2, real-valued function z of a vector variable (x,y), denote the partial derivative of z with respect to x as zx and the partial derivative of z with respect to y as zy.  Then the gradient of z is the vector grad z = (zx,zy).

Analogous to to the necessary condition that the first derivative vanish for critical points on a curve, a point (x0,y0) is a critical point if, and only if, grad z(x0,y0) = (0,0).

A critical point is a local or relative extremum if it is either a local minimum or a local maximum.  A critical point that is not a local extremum is called a saddlepoint, analogous to an inflection point on a curve.

Denote the partial derivatives of zx and zy as zxx, zxy = zyx, and zyy.  Then the Hessian matrix2 of z is the 2 x 2 symmetric matrix H(z) = [hi j(x,y)] where h11 = zxx, h12 = zxy = h21, and h22 = zyy.  The Hessian determinant of z is the discriminant D = |H| = h11 h22 - h122.

Analogous to the second derivative test for critical points on a curve, a critical point (x0,y0) is a

If D = 0 at (x0,y0), the critical point is called degenerate, and its classification requires further analysis.

The example presented in this lecture is solved using analytical techniques.  For a numerical example, see Newton's optimization method.


1Dept. of Civ. and Envir. Engrg., Univ. of South Florida, Tampa, FL 33620. [ Glossary | References ]
2After L. O. Hesse (1811-74), a student of C. G. J. Jacobi (1804-51). [ Printable Glossary ]

[ Main | Function | x-Partial | x-Zeros | y-Partial | y-Zeros | Solution | Hessian | Classification ]