I'm just trying to solidify the different cases for functions When dealing with functions that are $C^1$, say I have some function: $$f(a,b)=\begin{cases}(a^4+b^4)/(a^2+b^2)\quad &\text{ if } (a,b)\ne (0,0)\\ 0 &\text{ at }(0,0)\end{cases}$$

$C^1$ functions need to have their partials exist and be continuous everywhere in that neighbourhood. So to test if my function is continuously differentiable, I can first take the classical limit of the formula and test to see that the partials exist (that is, by fixing $y$, and adding arbitrary $h$ to $x$, and taking limit as $h\to0$ and vice versa for $y$ with $k\to0$).

Since they do in this case, do I then need to test if the partials $=0$ as $(a,b)\to(0,0)$ (taking the limit of the partials). Or would I take the limit of the function itself?

And if this function is continuously differentiable (which the limit does go to $0$, in both the original $f$, and $f'$ -- I'm not sure which one is more important), this function is differentiable everywhere, and thus $C^1$ ?


To check that such a function is $C^1$, you should

  1. Show that $\lim_{h\to 0}\frac{f(h,0)-f(0,0)}{h}$ and $\lim_{h\to 0}\frac{f(0,h)-f(0,0)}{h}$ exist. These are the partial derivatives at $(0,0)$; they must be computed by the definition because the usual limit theorems do not apply at $(0,0)$.

  2. Compute partial derivatives at points $(x,y)$ other than $(0,0)$. This is an exercise with the quotient rule.

  3. Show that the result of step 2 tends to the result of step 1 as $(x,y)\to (0,0)$.

You do not need to verify that $f$ itself is continuous, since you can invoke a theorem that says that continuous partials imply differentiability, and therefore continuity.

On the other hand, sometimes a function of this type turns out to be discontinuous at $(0,0)$, and then it's good idea to demonstrate that, thus showing that it's not $C^1$.

| cite | improve this answer | |
  • $\begingroup$ Thank you kindly. Clear and concise, just what I needed to sort the textbook's dense (although useful) and convoluted info. $\endgroup$ – sahimat Oct 3 '14 at 5:02

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.