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

I was looking at old complex analysis exams, and there is one problem I can't figure out.

"Use the partial fraction expansion of $\frac{z}{e^z-1}$ to show $\sum_1^\infty 1/k^2=\frac{\pi^2}{6}$."

I recognize that as the generating function for the Bernoulli numbers, but I think the point of the problem is to solve it "from scratch", without that kind of knowledge.

The function has simple poles at $2\pi i k$ for $k\in\mathbb{Z}$, and residue $2\pi i k$ at $2\pi i k$ . Unfortunately, the obvious series, with terms of the form $\frac{2\pi i k}{z-2\pi i k}$, doesn't converge. Adding convergence terms (like in the proof of Mittag-Leffler's theorem) I get a series with terms of the form $\frac{z^2}{z^2-k^2}$, modulo some constants, but I don't see where to go from there, because it vanishes at 0. I think point is that we are supposed to evaluate the partial fraction decomposition at 0, as the function is clearly 1 there.

Thanks for the help.

As noted in the comments, there is an answer here that looks similar to what is intended: , however it seems much too involved for an exam setting, and is deliberately not rigorous. Would it be possible to modify it to be simpler and faster?

Edit: I managed to figure out a fairly slick solution that's much better than the accepted answer. I don't have time to write it up right now. If you read this and want to see it, ping me by posting a comment to this question.

share|cite|improve this question
See Qiaochu's answer in the dupe: which answers this question, I believe. – Aryabhata Apr 17 '12 at 20:43
Qiaochu's answer isn't rigorous (the series he manipulates doesn't converge), and it seems like quite an involved answer to come up with on an exam. – Potato Apr 17 '12 at 20:49
I believe Qiaochu deliberately posted it "Euler style". As for too involved, depends on the course/professor etc etc. If you have questions about that answer perhaps you can try commenting on that answer there or editing your current question with a link to that saying that is not rigorous and how can it be made rigorous and simple etc. – Aryabhata Apr 17 '12 at 20:51
I have edited the question to address your concerns. – Potato Apr 17 '12 at 21:02
After the edit, I retract my close vote. – Aryabhata Apr 17 '12 at 21:04
up vote 2 down vote accepted

Potato, this is just an idea, but note that

$$\int\limits_0^\infty {\frac{x}{{{e^x} - 1}}dx} = \frac{{{\pi ^2}}}{6}$$

With a change of variables one has that

$$\int\limits_0^\infty {\frac{x}{{{e^x} - 1}}dx} = - \int\limits_0^1 {\frac{{\log \left( {1 - x} \right)}}{x}dx} $$

Now use

$$ - \frac{{\log \left( {1 - x} \right)}}{x} = \sum\limits_{k = 1}^\infty {\frac{{{x^{k - 1}}}}{k}} \text{ ; } |x|<1$$

from where

$$ - \int\limits_0^x {\frac{{\log \left( {1 - t} \right)}}{t}dt} = \sum\limits_{k = 1}^\infty {\frac{{{x^k}}}{{{k^2}}}} $$

This means that

$$\int\limits_0^\infty {\frac{x}{{{e^x} - 1}}dx} = - \int\limits_0^1 {\frac{{\log \left( {1 - t} \right)}}{t}dt} = \sum\limits_{k = 1}^\infty {\frac{1}{{{k^2}}}} $$

So you might want to use residues (which I don't know about), to calculate

$$\int\limits_0^\infty {\frac{z}{{{e^z} - 1}}dz} $$

since it has singularities at every $z_k=2\pi ki$

Another known approach is given here, starting at $(35)$

$$\frac{z}{2} + \frac{z}{{{e^z} - 1}} = \frac{z}{2}\coth \frac{z}{2}$$

from where

$$\frac{z}{2}\coth \frac{z}{2} = \sum\limits_{n = 0}^\infty {{B_{2n}}\frac{{{z^{2n}}}}{{\left( {2n} \right)!}}} $$

and then

$$z\coth z = \sum\limits_{n = 0}^\infty {\frac{{{2^{2n}}{B_{2n}}}}{{\left( {2n} \right)!}}{z^{2n}}} $$

they then let $z=iz$, from where

$$z\cot z = \sum\limits_{n = 0}^\infty {{{\left( { - 1} \right)}^n}\frac{{{2^{2n}}{B_{2n}}}}{{\left( {2n} \right)!}}{z^{2n}}} $$

They go on with partial fractions expansions, but I remember seeing elsewhere this:

$$\sin z = z\prod\limits_{n = 1}^\infty {\left( {1 - \frac{{{z^2}}}{{{n^2}{\pi ^2}}}} \right)} $$


$$\log \sin z = \sum\limits_{n = 1}^\infty {\log \left( {1 - \frac{{{z^2}}}{{{n^2}{\pi ^2}}}} \right)}+\log z $$

Differentiating and multiplying by $z$ gives

$$z\cot z = 1-2\sum\limits_{n = 1}^\infty {\dfrac{{ \dfrac{{{z^2}}}{{{n^2}{\pi ^2}}}}}{{1 - \dfrac{{{z^2}}}{{{n^2}{\pi ^2}}}}}} $$

but since

$$\frac{{\frac{{{z^2}}}{{{n^2}{\pi ^2}}}}}{{1 - \frac{{{z^2}}}{{{n^2}{\pi ^2}}}}} = \sum\limits_{k = 1}^\infty {\frac{1}{{{n^{2k}}{\pi ^{2k}}}}} {z^{2k}}$$

they write

$$z\cot z = 1- 2\sum\limits_{n = 1}^\infty {\sum\limits_{k = 1}^\infty {\frac{1}{{{n^{2k}}}}} \frac{{{z^{2k}}}}{{{\pi ^{2k}}}}} $$

then changing the order of the sum

$$z\cot z = 1 - 2\sum\limits_{k = 1}^\infty {\frac{{\zeta \left( {2k} \right)}}{{{\pi ^{2k}}}}{z^{2k}}} $$

Thus, since

$$z\cot z = 1 + \sum\limits_{n = 1}^\infty {{{\left( { - 1} \right)}^n}\frac{{{2^{2n}}{B_{2n}}}}{{\left( {2n} \right)!}}{z^{2n}}} $$

they argue that

$${\left( { - 1} \right)^{n + 1}}\frac{{{2^{2n - 1}}{B_{2n}}}}{{\left( {2n} \right)!}}{\pi ^{2n}} = \zeta \left( {2n} \right)$$

share|cite|improve this answer
Thanks for your opinion. However I deleted it definitely due to a downvote. +1 for yours. – Américo Tavares Apr 17 '12 at 23:46
@AméricoTavares Mine was downvoted too. I guess someone is not very happy with these contributions. I made it a CW because: 1. It doesn't fully answer the question, but is rather a very long comment-hint. 2. Anyone can add more info, correct an error, or provide a better hint. If someone is not happy with the contribution, comment and/or edit. Downvoting is rather pointless, IMO. – Pedro Tamaroff Apr 17 '12 at 23:50
I've seen the downvote in yours. When I undeleted mine I made it a CW too, having seen yours was a CW. – Américo Tavares Apr 17 '12 at 23:54

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.