Multivariable calculus

Second Partial Derivatives of x/(1 + y^2)

For f(x,y) = x/(1 + y^2) the second partials are f_xx = 0, f_yy = x(6y^2 - 2)/(1 + y^2)^3, and the mixed partial f_xy = -2y/(1 + y^2)^2. The pure x partial vanishes because the function is linear in x, but the surface is still curved: the mixed partial is nonzero whenever y is not zero.

f(x,y)=x1+y2f(x,y) = \frac{x}{1 + y^{2}}
fx=11+y2\frac{\partial f}{\partial x} = \frac{1}{1 + y^{2}}
fy=2xy(1+y2)2\frac{\partial f}{\partial y} = \frac{-2xy}{\left(1 + y^{2}\right)^{2}}

Second and mixed partials

2fx2=0\frac{\partial^{2} f}{\partial x^{2}} = 0
2fyx=2y(1+y2)2\frac{\partial^{2} f}{\partial y\,\partial x} = \frac{-2y}{\left(1 + y^{2}\right)^{2}}
2fy2=x(6y22)(1+y2)3\frac{\partial^{2} f}{\partial y^{2}} = \frac{x\left(6y^{2} - 2\right)}{\left(1 + y^{2}\right)^{3}}

Splitting the work between the two variables

In xx this function is linear: the whole factor 11+y2\frac{1}{1 + y^{2}} is a constant slope. In yy it is a rational function with a denominator that never vanishes, so both partials exist everywhere.

fx=11+y2,fy=2xy(1+y2)2\frac{\partial f}{\partial x} = \frac{1}{1 + y^{2}}, \qquad \frac{\partial f}{\partial y} = \frac{-2xy}{\left(1 + y^{2}\right)^{2}}

Differentiating fxf_x again in xx gives zero, since 11+y2\frac{1}{1 + y^{2}} has no xx in it. Differentiating fxf_x in yy instead gives the mixed partial through the power rule on (1+y2)1\left(1 + y^{2}\right)^{-1}.

2fx2=0,2fyx=2y(1+y2)2\frac{\partial^{2} f}{\partial x^{2}} = 0, \qquad \frac{\partial^{2} f}{\partial y\,\partial x} = \frac{-2y}{\left(1 + y^{2}\right)^{2}}

For fyyf_{yy}, differentiate 2xy(1+y2)2-2xy\left(1 + y^{2}\right)^{-2} using the product rule, then factor out one power of the denominator.

2fy2=2x(1+y2)4y2(1+y2)3=x(6y22)(1+y2)3\frac{\partial^{2} f}{\partial y^{2}} = -2x\,\frac{\left(1 + y^{2}\right) - 4y^{2}}{\left(1 + y^{2}\right)^{3}} = \frac{x\left(6y^{2} - 2\right)}{\left(1 + y^{2}\right)^{3}}

The mistake students make

Seeing fxx=0f_{xx} = 0, students often conclude that the surface has no curvature at all, or that the Hessian is degenerate in a way that makes the second derivative test unnecessary. Neither follows. The Hessian determinant here is D=0fyy(fxy)2=(2y(1+y2)2)2D = 0 \cdot f_{yy} - \left(f_{xy}\right)^{2} = -\left(\frac{2y}{(1 + y^{2})^{2}}\right)^{2}, which is negative whenever y0y \neq 0.

What fxx=0f_{xx} = 0 really says is narrower: every cross section taken at fixed yy is a straight line, so the surface is ruled. Along other directions it bends, and the mixed partial is what records that bending.

The second slip is a chain rule miss on the way to fyyf_{yy}. Differentiating (1+y2)2\left(1 + y^{2}\right)^{-2} produces 2(1+y2)32y-2\left(1 + y^{2}\right)^{-3} \cdot 2y, and dropping that inner 2y2y leaves a formula whose sign is wrong for large yy.

Clairaut symmetry with an asymmetric function

This function is not symmetric in xx and yy, which makes it a better test of Clairaut's theorem than a symmetric example. Start from fyf_y and differentiate in xx.

2fxy=x(2xy(1+y2)2)=2y(1+y2)2\frac{\partial^{2} f}{\partial x\,\partial y} = \frac{\partial}{\partial x}\left(\frac{-2xy}{\left(1 + y^{2}\right)^{2}}\right) = \frac{-2y}{\left(1 + y^{2}\right)^{2}}

That matches the other order, even though the two computations look nothing alike: one was a power rule in yy, the other a constant multiple rule in xx. Continuity of both mixed partials is what forces the agreement, and here the denominator is at least 1 everywhere, so continuity holds on the whole plane.

The mixed partial is negative for y>0y > 0 and positive for y<0y < 0, with extreme values at y=±13y = \pm\frac{1}{\sqrt{3}}. That is the rate at which the slope in xx changes as you move in yy, which is exactly what a mixed partial measures.

Frequently asked questions

Does f_xx = 0 mean the graph is a plane?

No. It means each slice at fixed yy is a straight line. The lines change slope as yy changes, and the surface bends in the yy direction, so the graph is a curved ruled surface rather than a plane.

Where does f_yy change sign?

Where 6y2=26y^{2} = 2, that is y=±13±0.577y = \pm\frac{1}{\sqrt{3}} \approx \pm 0.577. The sign also depends on xx, since fyyf_{yy} carries a factor of xx and flips as xx crosses zero.