Multivariable calculus
Partial Derivatives of x^2 sin(y)
For f(x, y) = x squared times sin(y), the partial derivative with respect to x is 2x sin(y), and the partial derivative with respect to y is x squared times cos(y). Each partial differentiates only its own factor and keeps the other one as a constant multiplier.
One factor moves, the other is a number
Hold fixed. Then is some number , and is a plain parabola in with derivative . Restore .
Hold fixed instead. Now is a number and is a sine wave in scaled by that number, so the derivative is .
Read them at : and . The surface is at a crest in the direction there, so moving in costs nothing to first order.
The mistake: mixing the two rules together
A frequent wrong answer is , which differentiates both factors in the same step. Another is , which is the full product rule as if and were the same variable. Both come from not deciding which variable is moving.
- differentiates only, correct for .
- differentiates only, correct for .
- The sum of the two is the rate of change along the direction , not a partial derivative at all.
If you ever need that sum, it has a name: it is , the rate at which changes along a path that moves one unit in for every unit in . Divide it by if you want the directional derivative in the unit direction, since has length .
Second derivatives and the shape
Differentiate again to see the curvature. and . Off the line , wherever the surface curves upward in and downward in , so those points sit on saddle shaped patches. On itself the curvature is zero, and in fact both partials vanish along that whole line, so the second derivative test says nothing there.
The mixed partial is whichever order you take, since differentiating by and by both land there.
Frequently asked questions
Where do both partials vanish?
You need and . If both are zero for every , so the whole axis is a line of critical points. If you would need and zero at once, which never happens.
Does the answer change if I write the function as sin(y) x^2?
No. Multiplication commutes, and each partial derivative only cares about which factor contains the moving variable.