I'm trying to compute the directional derivative at $(0,0)$ of $f$ in the direction $u=(u_1\ \ u_2)^T$ with $||u|| = 1$, where $f(x,y) =\begin{cases} \frac{x^3}{x²+y²} & (x,y) \neq (0,0) \\ 0 & (x,y) = (0,0)\\ \end{cases}$

Now, I computed the partial derivatives of f at $(0,0)$, which I found to be respectively $1$ and $0$. Thus, I would expect the directional derivative in the direction $u=(u_1\ \ u_2)^T$ to be $(1 \ \ 0) \cdot (u_1\ \ u_2)^T = u_1$. However, according to the answer sheet, the answer should be $u_1^3$. What am I doing wrong?


What you're doing wrong is assuming that $f$ is smoothly differentiable at $(0,0)$. Instead of using the gradient shortcut to calculate the directional derivative, you should just use the definition of directional derivatives. (By the way, as a practical note, you haven't used the hypothesis that $\|u\| = 1$.)

I'll remind you of that definition. For convenience's sake, write $f$ as a function of vectors $(x,y) = \mathbf{x}$. The directional derivative of $f$ at $\mathbf{x}$ in the direction of $\mathbf{v}$ is $$\lim_{t\to 0}\frac{1}{t} \bigg( f(\mathbf{x} + t\mathbf{v}) - f(\mathbf{x})\bigg).$$ So now do the only thing you can: plug in for $\mathbf{x}$, $\mathbf{v}$, and compute the limit.

  • 1
    $\begingroup$ Thanks! Somehow that definition isn't in my notes or book, so I didn't think of using it (though I probably should have thought of looking up the definition on the internet). $\endgroup$ – Max Feb 15 '13 at 4:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.