Multivariable calculus

Critical Points of x^2 + y^2 + x^2 y

f(x,y) = x^2 + y^2 + x^2 y has three critical points. The origin is a local minimum with discriminant D = 4 and f_xx = 2. The points (sqrt 2, -1) and (-sqrt 2, -1) are saddles with D = -8. The general discriminant is D = 4 + 4y - 4x^2, which is what separates the three cases.

f(x,y)=x2+y2+x2yf(x,y) = x^{2} + y^{2} + x^{2}y
  • (0,0)\left(0, 0\right) local minimumdiscriminant D = 4
  • (1.4142135623730951,1)\left(1.4142135623730951, -1\right) saddle pointdiscriminant D = -8
  • (1.4142135623730951,1)\left(-1.4142135623730951, -1\right) saddle pointdiscriminant D = -8

Factor the partial that factors

fx=2x+2xy=2x(1+y),fy=2y+x2\frac{\partial f}{\partial x} = 2x + 2xy = 2x(1 + y), \qquad \frac{\partial f}{\partial y} = 2y + x^{2}

The xx equation splits into two cases, and each has to be followed through separately.

  • Case x=0x = 0: then 2y=02y = 0, so y=0y = 0 and the point is (0,0)(0,0).
  • Case y=1y = -1: then 2+x2=0-2 + x^{2} = 0, so x=±2x = \pm\sqrt{2} and the points are (2,1)\left(\sqrt{2},-1\right) and (2,1)\left(-\sqrt{2},-1\right).

Three critical points, and no others, since the two cases exhaust the solutions of 2x(1+y)=02x(1+y) = 0.

Classifying all three

fxx=2+2y,fyy=2,fxy=2xf_{xx} = 2 + 2y, \qquad f_{yy} = 2, \qquad f_{xy} = 2x
D(x,y)=2(2+2y)4x2=4+4y4x2D(x,y) = 2(2 + 2y) - 4x^{2} = 4 + 4y - 4x^{2}

At the origin, D=4>0D = 4 > 0 and fxx=2>0f_{xx} = 2 > 0, so it is a local minimum with f(0,0)=0f(0,0) = 0. At (±2,1)\left(\pm\sqrt{2},-1\right) the term 4y4y cancels the leading 44 and 4x2=84x^{2} = 8 remains, so D=8<0D = -8 < 0 and both are saddles, each at height f=1f = 1.

The mistake students make

Dividing 2x(1+y)=02x(1+y) = 0 by 2x2x throws away the origin, and stopping after x=0x = 0 throws away both saddles. A factored equation is a signal to branch, not to cancel: every factor gets its own case, and each case is then substituted into the other equation.

The second slip is the sign of fxxf_{xx} at the saddles. There fxx=2+2(1)=0f_{xx} = 2 + 2(-1) = 0, which looks alarming, but DD is already negative and that alone settles the classification. The sign of fxxf_{xx} is consulted only when D>0D > 0.

Frequently asked questions

Is the origin a global minimum?

No, only a local one. Fix y=10y = -10 and the function becomes x2+10010x2=1009x2x^{2} + 100 - 10x^{2} = 100 - 9x^{2}, which runs to -\infty as xx grows. For instance f(10,10)=800f(10,-10) = -800, well below f(0,0)=0f(0,0) = 0.

Why are the two saddle points at the same height?

Every xx in the formula appears as x2x^{2}, so f(x,y)=f(x,y)f(-x,y) = f(x,y) and the surface is a mirror image across the yy axis. That reflection swaps the two saddle points, forcing the same height f=1f = 1 and the same discriminant D=8D = -8.