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

The integral I want to evaluate is the integral of $\displaystyle\frac{z^n}{e^z-1}$ over the circle $|z| = 3 \pi $. I was thinking of using Cauchy's Residue Theorem, but then I got stuck on evaluating the residues at $z=0$, because I couldn't figure out how to change the form into one where I could find a Laurent series.

share|cite|improve this question
One formula you should be remembering for your test (it's probably the most useful one for calculating residues) is that $\operatorname{Res}_{z_{0}} \frac{f}{g} = \frac{f(z_0)}{g'(z_{0})}$ if $f(z_{0}) \neq 0$ and $g(z_{0}) = 0$ and $g'(z_{0}) \neq 0$. (i.e. $z_{0}$ is a simple zero of $g$). You can find other useful ones on the Wikipedia page on the residue theorem. – t.b. May 13 '11 at 7:29
up vote 7 down vote accepted

OK, this is an attempt to rewrite my answer without mentioning L'Hôpital or Riemann, as requested.

The integrand $f(z) = \frac{z^n}{\mathrm e^z-1}$ is the quotient of two holomorphic functions, each of which we can write as a power series at any point $z_0$. For $z_0=0$, this yields

$$f(z)=\frac{z^n}{\left(\sum_{k=0}^\infty z^k/k!\right)-1}=\frac{z^n}{\sum_{k=1}^\infty z^k/k!}=\frac{z^n}{z\sum_{k=1}^\infty z^{k-1}/k!}=\frac{z^n}{z\left(1+\sum_{k=2}^\infty z^{k-1}/k!\right)}\;.$$

If $n$ is a positive integer, we can cancel one factor $z$ to obtain

$$f(z)=\frac{z^{n-1}}{1+\sum_{k=2}^\infty z^{k-1}/k!}\;.$$

The denominator doesn't go to zero as $z\to0$, and thus there is no pole at $z=0$.

On the other hand, if $n=0$, we have

$$f(z)=\frac{1}{z\left(1+\sum_{k=2}^\infty z^{k-1}/k!\right)}\;.$$

Thus we can determine

$$\lim_{z\to0}zf(z) = \lim_{z\to0}\frac{1}{\left(1+\sum_{k=2}^\infty z^{k-1}/k!\right)}=1\;,$$

and this is the residue of $f$ at $z=0$.

The residues at $z=\pm2\pi\mathrm i$ can be determined similarly using $\mathrm e^{z}=\mathrm e^{z\mp2\pi\mathrm i}$.

share|cite|improve this answer
I wouldn't call this l'Hôpital but rather Riemann's theorem on removable singularities (Riemannscher Hebbarkeitssatz), but the conclusion remains the same, of course. – t.b. May 12 '11 at 23:34
@joriki Two other poles enter the picture, at $\pm2\pi\mathrm{i}$. Their contributions to the residue formula add up when $n$ is even and cancel each other when $n$ is odd. – Did May 12 '11 at 23:40
@I Love Cake: Do you know about Laurent series? If $f$ only has a simple pole at $z=a$, then the only negative power in the Laurent series centered at $a$ is the -1 power. This means that we can write $f(z) = c_{-1} \frac{1}{(z-a)} + \sum_{n=0}^\infty c_n (z-a)^n$. The residue of $f$ at $a$ is going to be the coefficient $c_{-1}$, so one way of recovering this is by multiplying $f(z)$ by $z-a$ and taking the limit as $z \to a$, since the nonnegative powers in the sum will vanish. Does that make sense? – user1736 May 13 '11 at 0:36
@Theo: I think you can actually derive L'hopital's rule for analytic functions. If $f$ has a zero of order $m$ and $g$ has a zero of order $n$, one way of finding the limit of the quotient is to factor out the zeros and see what cancels out. Alternatively, you can also differentiate $f$ and $g$ that many times directly, so that you eventually end up with the same expression. I guess you can make the argument that this means that you don't really ever need l'hopital's rule, but it should still be valid. – user1736 May 13 '11 at 0:43
@user: Ok, very good. Actually, I think you should elaborate your comments into an answer because this very much looks like what I Love Cake is looking for (I'm too tired now and I've written too many answers, recently) – t.b. May 13 '11 at 1:18

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.