Multivariable calculus
Partials of (x^2 - y^2)/(1 + x^2 y^2)
For f(x, y) = (x^2 - y^2)/(1 + x^2 y^2) the partial with respect to x is 2x(1 + y^4)/(1 + x^2 y^2)^2, and the partial with respect to y is -2y(1 + x^4)/(1 + x^2 y^2)^2. The cross terms cancel during simplification, which is why the answers are far tidier than the raw quotient rule output.
Apply the quotient rule, then collect terms
Hold fixed. The numerator has derivative , and the denominator has derivative .
Factor out of the numerator and expand what is left: . The terms cancel exactly.
The same computation in gives over the same squared denominator, with the minus sign inherited from the in the numerator. At the two partials are and .
The mistake: stopping before the cancellation
The unsimplified quotient rule output is technically correct, but leaving it there hides the structure and makes the next step, a critical point hunt or a second derivative, far harder than it needs to be.
- Expand the numerator fully before you decide it cannot be simplified.
- Here appears twice with opposite signs and disappears.
- The clean form shows immediately that has the same sign as , since always.
The other frequent error is a sign: , and forgetting that minus flips the whole answer.
What the signs tell you
Because and are always positive and the denominator is squared, the signs are decided entirely by and .
- when and when .
- when and when .
- Both vanish only at the origin, the single critical point.
That is exactly saddle behaviour: moving away from the origin along the axis increases , and moving away along the axis decreases it. Restricted to the axes, behaves like in one direction and in the other.
The denominator is at least , so the function and both partials are defined and smooth on the entire plane, with no excluded curves to state.
Frequently asked questions
Why did the x^2 y^2 terms cancel?
The quotient rule numerator is . Expanding the second product gives , and the exactly cancels the from the first product, leaving .
Is the origin a saddle point?
Yes. The second partials at the origin are , , and , so the discriminant is . The surface rises along the axis and falls along the axis from that point.