Multivariable calculus
Partial Derivatives of arctan(x + yz)
For f(x,y,z) = arctan(x + yz), the partial derivative with respect to x is 1 over 1 plus (x + yz) squared. The y partial is z times that, and the z partial is y times that. The arctangent supplies one common denominator and the chain rule supplies the numerators 1, z and y.
One outside derivative, three inside derivatives
Let , so . The derivative of is times the derivative of . Since is the same for all three variables, the denominator is identical in all three answers, and only the inside derivative changes.
Differentiate one variable at a time. With respect to the term is constant, so . With respect to the is constant and behaves like a constant times , so . Symmetrically .
So the three answers are the same fraction with numerators , and . Compute the denominator once and reuse it, which is faster and removes a chance to make three different transcription errors.
The mistake: squaring the wrong thing
The denominator is where is the entire inside expression, so it is and not . Those differ by the cross term , which is not zero in general. Keep the parentheses in place through the whole calculation.
The second slip is swapping the numerators of the and partials. Differentiating with respect to leaves behind, not . The general rule holds: the variable you differentiated should not survive in the numerator that the chain rule produces.
- Wrong denominator: , which drops the cross term.
- Wrong numerator: for the partial, when it should be .
- Right: one denominator , numerators , and .
Structure and a sample point
Because the denominator is shared, the other two partial derivatives are simple multiples of the first: and . That relation is a fast internal consistency check on an answer you have just written.
At the inside is , so the denominator is and all three partial derivatives equal . Note also that the denominator is at least everywhere, so can never exceed in size. The other two are not bounded, because they carry the extra factors and : at the denominator is and .
Frequently asked questions
Where is the x partial derivative largest?
Wherever , since that makes the denominator as small as possible, namely . There , which is the steepest the arctangent ever gets in the direction.
Can the y partial derivative be negative?
Yes. The denominator is always positive, so the sign of is the sign of . It is negative whenever and zero on the plane .