Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

There is a standard theorem that says if $f$ is analytic in the whole complex plane and never zero, then it has the form $e^g$ where $g$ is analytic in the whole plane; i.e. we can define a function $\log f$ that is analytic in the whole plane and satisfies $f=e^{\log f}$. The argument is easy: $f$ is entire and never zero, so $f'/f$ is entire, so $g(z)=\int_0^z f'(\zeta)d\zeta/f(\zeta)$ is well-defined and entire, and then computation shows that $fe^{-g}$ has derivative zero, so is constant, so $f$ is a constant multiple of $e^g$, and the constant can be absorbed into $g$.

The argument does not mention the topology of the situation, and this bugs me a little. By the Little Picard theorem, the image of $f$ is the whole complex plane minus zero. In general, of course, a continuous single-valued branch of $\log$ cannot be defined on this set. That's why the proof involved the integral of the logarithmic derivative of $f$, rather than merely the composition of $f$ with the right branch of $\log$. That said, if we obtained a different function $h$ by picking some branch of $\log$ defined on $\mathbb{C}$ minus some ray from the origin (call this set $S$), and then defining $h(z)=\log f(z)$ on $f^{-1}(S)$, then of course $h$ would satisfy $f=e^h$ where it is defined, and then $1=e^{h-g}$, so that on each connected component of $f^{-1}(S)$, since $h,g$ are defined and continuous they must differ by a constant multiple of $2\pi i$. If I allow myself to use a different branch of $\log$ for each connected component of $f^{-1}(S)$, I could end up with $h$ exactly equaling $g$ and therefore being extendable by continuity to all of $\mathbb{C}$. What this adds up to is that I have the strong feeling that the theorem (in spite of the easy calculation-based proof) has some nontrivial topological content and I am trying to isolate this content.

At this point, I was originally going to ask for help isolating the topological content. But over the course of composing the question, I think I may have answered it. So, following is what I believe to be an alternative proof of the original result that makes the topological content explicit. My question is now, is the following proof correct? I would also be interested in your thoughts about (soft-question) where is the topology hiding in the original proof?

Theorem: If $f$ is entire and never zero it has the form $f=e^g$ for $g$ entire.

Proof: Let $f$ be an entire function that is never zero. Then $f:\mathbb{C}\rightarrow\mathbb{C}\setminus\{0\}$ is continuous, so induces a homomorphism $f_*:\pi_1(\mathbb{C})\rightarrow\pi_1(\mathbb{C}\setminus\{0\})$, which is trivial because $\mathbb{C}$ is simply connected. Therefore, $f$ lifts to a continuous function $\hat{f}$ from $\mathbb{C}$ to $\mathbb{C}\setminus{0}$'s universal cover $U$. But $U$ is the Riemann surface of the $\log$ function; in particular, the logarithm is well-defined and single-valued on $U$; so let $g=\log \hat{f}$; and then $g$ is such that $f=e^g$. $\hat{f}$ is actually holomorphic to $U$, so the $g$ thus defined is entire.

EDIT: When I wrote this proof I was a little uncomfortable with speaking of $\hat{f}$ being holomorphic to $U$ and the log being holomorphic on $U$, since my understanding of covering space theory is strictly topological. Subsequently I realized that we can actually take $U$ to be $\mathbb{C}$, with the covering map given by $z\mapsto e^z$, and then the "logarithm on $U$" is just the identity on $\mathbb{C}$. I think the argument goes through, but can be made even shorter:

By the lifting theorem, any function $f$ from a simply connected domain to $\mathbb{C}\setminus\{0\}$ has a lift $\hat{f}$ to $\mathbb{C}\setminus{0}$'s universal cover, $\mathbb{C}$, that factors through the covering map $z\mapsto e^z$. Then $f=e^{\hat{f}}$ and $\hat{f}$ is the desired $g$. Since the covering map is holomorphic, $f$ and $\hat{f}$ are simultaneously holomorphic. Done!

Thanks to Jonas Meyer for highlighting that the key feature of the original argument is also the simple-connectedness of the domain, on which rests the invocation of Cauchy's theorem to show the integral is well-defined.

share|cite|improve this question
up vote 11 down vote accepted

Where is the topology hiding in the original proof?

It is hiding in the phrase "is well-defined," or, equivalently, in Cauchy's theorem. In a domain that is not simply connected, this "definition" of $g$ in general depends on the path taken from your base point to $z$. (For example, if there are isolated singularities, then different paths could wind around the singularities differently and pick up residues.) The well-definedness depends on Cauchy's theorem for simply connected domains.

share|cite|improve this answer
Ah. Your point is that as soon as $f$ is analytic and nonvanishing (forget entire) in a simply-connected domain $D$, it can be written as $e^g$ on $D$, and the original argument goes through to prove this? (I.e. that really the content of "$f$ is entire" was "f is analytic in a simply connected domain," and the scope of the theorem applies to whatever this domain is?) – Ben Blum-Smith Dec 6 '11 at 23:03
Yes. And this characterizes simply connected domains in $\mathbb C$. The direction we're discussing follows from Cauchy's theorem for simply connected domains. Conversely, if $G$ is a connected open subset of $\mathbb C$ such that for every nonvanishing analytic $f:G\to\mathbb C$ there exists an analytic $g:G\to\mathbb C$ such that $f=e^g$, then $G$ is simply connected. This is made clear in Conway's Functions of one complex variable, where the only property about a simply connected domain used to prove the Riemann Mapping Theorem is exstnce of analytic sqrts of nonvnshng analytic fnctns. – Jonas Meyer Dec 6 '11 at 23:21
Good comment. Excellent. – mick Jul 13 '13 at 23:12

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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