In the real numbers the expression $x^x$ converges to 1 coming from "any" direction (which here means from the positive side or from the negative side). Now I feel to remember that this does not hold in the complex plane, because there the expression $z^z$ should be able to take any value (when $|z|$ goes to 0) depending on the direction/angle you're coming from.

Now I tried to make sense of that, but couldn't complete the picture. I will demonstrate how far I got. Let $z = r \cdot e^{i \phi}$ then

$z^z = (r \cdot e^{i \phi})^{r \cdot e^{i \phi}} = (r)^ {r \cdot e^{i \phi}} \cdot (e^{i \phi})^{r \cdot e^{i \phi}} = (r)^ {r \cdot e^{i \phi}} \cdot e^{i \phi \cdot (r \cdot e^{i \phi})}$.

With this parametrization, when we look at $|z| \rightarrow 0$, this just means $r \rightarrow 0$. My intuition (and Wolframalpha) would now say that the term with $r$ as the base would approach 1 as $r$ approaches 0. But now Wolframalpha also says that the second term with $e$ as the base also approaches 1 as $r$ approaches 0, which would mean the combined limit is 1, not depending on $\phi$. This would then mean that $z^z$ does indeed also approach 1 in the complex plane from whereever we are coming (as the limit is independent of $\phi$).

Now, where is the flaw?

  1. Does $z^z \rightarrow 1$ as $|z| \rightarrow 0$ for all $z \in \mathbb{C}$?
  2. Is there a flaw in my calculations? (either in an intermediate step or the reasoning at the end)
  • 1
    $\begingroup$ Are you certain that $(r \cdot e^{i \phi})^{r \cdot e^{i \phi}} = (r)^ {r \cdot e^{i \phi}} \cdot (e^{i \phi})^{r \cdot e^{i \phi}}$? This seems like a dangerous step to me. $\endgroup$
    – Arthur
    Aug 17 '17 at 13:22
  • $\begingroup$ $(a \cdot b)^c = a^c \cdot b^c$ does hold in the complex plane as well, so I'd say yes. $\endgroup$
    – tehfurbolg
    Aug 17 '17 at 13:23
  • 2
    $\begingroup$ Even with both exponents and bases being complex? $\endgroup$
    – Arthur
    Aug 17 '17 at 13:25
  • 2
    $\begingroup$ @tehfurbolg You may find scipp.ucsc.edu/~haber/ph116A/clog_11.pdf useful to consult, and avoid false beliefs about the complex plane :) $\endgroup$ Aug 17 '17 at 13:28
  • 2
    $\begingroup$ When we say "$0^0$ is an indeterminate form" we do not mean that $z^z$ may converge to different values as $z \to 0$. Instead we mean $f(z)^{g(z)}$may converge to different values for functions $f,g$ with $f(z) \to 0, g(z) \to 0$ as $z\to 0$. $\endgroup$
    – GEdgar
    Aug 17 '17 at 14:31

First of all, $z^z$ is not analytic at $z=0$. It is defined as $z^z=e^{z\log z}$, where $\log$ is an analytic branch of the logarithm. Take the usual branch, defined on $\Bbb C\setminus(-\infty,0]$ with $\log1=0$. If $z=r\,e^{i\theta}$, $-\pi<\theta<\pi$, then $$ |z\log z|=r\,|\log r+i\,\theta|\le r(|\log r|+\pi), $$ which goes to $0$ as $r\to0$. It follows that $z^z\to1$ as $z\to0$.


The problem with the function $z\mapsto z^z=e^{z\ln z}$ is that it is not uniquely defined because the complex logarithm $\ln z$ isn't. However, if you fix a path in the complex plane, there is indeed a way to define $z^z$ uniquely on this path. Now if you fix path going to zero, you can ask if the limit of $z^z$ is indeed 1 along this path.

As @Julian Aguirre shows, if this path is in the subspace $\mathbb{C}\setminus (-\infty;0]$, then $z^z$ has a limit and it is $1$. However, there are paths where $z^z$ has no limits.

Let $z=re^{i\theta}$ so that $\ln z=\ln r + i\theta + 2ik\pi$. Then $$ z^z=e^{z\ln z}=e^{z(\ln r+\theta+2ik\pi)}$$ Now let's have a look at the exponent.

When $|z|\rightarrow 0$, $z\ln r\rightarrow 0$ and $zi\theta\rightarrow 0$. However $z\times 2ik\pi$ might not have a limit because you don't know how $k$ behave.

Concretely, we will fix a path in $\mathbb{C}$ going to $0$ and compute $z^z$ along this path. In fact, instead of $z^z$, we will look at $z\ln z$ which is simpler. We will show that there is some path where there is no limits.

The path is the following : $z=t\times e^{i/t}$. (Draw a picture of that path, this is spiral).

Then $\ln z=\ln t+ \frac{i}{t}$ and $z\ln z=z\ln t+ z\times \frac{i}{t}=z\ln t+i e^{i/t}$.

But then, you see that $z\ln z$ has no limit along this path. And neither has $z^z$.


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.