The Discriminant D
You'll learn: How to compute the discriminant D from the Hessian matrix to classify critical points of functions in two variables.
What Is the Discriminant?
After finding critical points of a multivariable function and computing the Hessian matrix (all those second-order partial derivatives), you need a single number that tells you what type of critical point you've found. That's where the discriminant comes in.
The discriminant D is a special combination of the Hessian matrix entries that acts like a fingerprint for the critical point's behavior.
Computing the Discriminant
For a function f(x, y), the discriminant D at a critical point is:
D = f_xx · f_yy - (f_xy)²
Notice this formula:
- Multiplies the two "pure" second derivatives (f_xx and f_yy)
- Subtracts the square of the mixed partial derivative (f_xy)²
Think of D as measuring whether the curvatures in the x- and y-directions "agree" or "fight" with each other. When they agree strongly, D is positive (you get a local max or min). When they fight, D is negative (you get a saddle point). The mixed derivative term f_xy represents twisting, which can disrupt agreement.
Example
If f_xx = 6, f_yy = 4, and f_xy = 1, then:
D = (6)(4) - (1)² = 24 - 1 = 23
This positive value suggests the critical point might be a local extremum (you'd still check f_xx to determine max vs. min).
Key Takeaway: The discriminant D = f_xx·f_yy - (f_xy)² condenses the Hessian's information into one number that helps classify whether a critical point is a local maximum, local minimum, or saddle point.