Take the function $f(x,y) = x^y$ where $x \in [0,\infty)$ and $y \in (-\infty,\infty)$.
The value of $f(0,0)$ is indeterminate. I want to know:

Is there some path $S$ on the $(x,y)$-plane that ends at the origin, $(0,0)$, such that $$\lim_{S\to(0,0)} x^y \to\infty$$

And if not, how do you prove that there isn't one?

It's trivial to show that there are limits that can equal $0$ or $1$:

For $x = 0$, $\lim\limits_{y \to 0} \ 0^y = 0 $, and for $y = 0$, $\lim\limits_{x \to 0} \ x^0 = 1 $.

However, I suspect there could be a limit approaching $\infty$ because,

If you take some $\delta < 0$ that may be arbitrarily close to $0$, then

$$ \lim_{x \to 0} x^{\delta} \to \infty$$

Therefore, in all neighbourhoods centred at $(0,0)$ there exist points at which $x^y$ is arbitrarily large (or approaching infinity if that can be claimed of a point).

It seems, to me, that if there's a path for which $x \to 0$ faster than $y \to 0$ than the value of $x^y$ could keep rising as the path goes on and the limit approaching the origin will approach infinity.


As you say, a small positive $x$ raised to a very small negative $y$ will be a large positive number. How small must $y$ be? If we want $x^{y}=l$, we simply set $y=\ln(l)/{\ln(x)}$.

Now, define parametric functions $x$ and $y$ by $x(t)=\frac{1}{t}$ and $y(t)=\ln(\ln(t))/\ln(x(t))=\ln(\ln(t))/\ln\left(\frac{1}{t}\right)$. Then the path $$ S(t)=\langle x(t),y(t)\rangle =\left\langle \frac{1}{t},\frac{\ln(\ln(t))}{\ln\left(1/t\right)}\right\rangle$$ has the property that it approaches $\langle 0,0\rangle$ as $t\to\infty$, but the value of $f\circ S(t)$ reduces to $\ln(t)$ which clearly increases without bound.

  • $\begingroup$ So, any path $S(t) =\left\langle \frac{1}{t},\frac{\ln(g(t))}{\ln\left(1/t\right)}\right\rangle$ where $g(t)$ is some increasing function such that as $t \to \infty \implies g(t) \to \infty$ works, right? Does that capture the entire family of solutions? $\endgroup$ – XYZT Mar 2 '14 at 6:21
  • 1
    $\begingroup$ I think so, except for two caveats. Firstly, you only need $\lim_{t\to\infty}g(t)=\infty$; it doesn't need to be monotone. Secondly, you need to make sure that $g$ doesn't increase too fast, since you need $$y(t)=\frac{\ln(g(t))}{\ln(1/t)}$$ to converge to $0$. $\endgroup$ – Unwisdom Mar 2 '14 at 6:25
  • $\begingroup$ Thanks! That was a great answer. Elegantly solved (in my opinion). $\endgroup$ – XYZT Mar 2 '14 at 6:32

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.