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

Today I have encounter a series:

$$\sum_{n=-\infty}^{+\infty}\frac{1}{(u+n)^2}=\frac{\pi^2}{(\sin \pi u)^2}$$ where $u \not \in \Bbb{Z}$ . I have known a method to computer it (by Residue formula): $$\int_{|z|=N+1/2}\frac{\pi \cot \pi z}{(u+z)^2}\text{dz}=-\frac{\pi^2}{\sin^2 \pi u}+\sum_{k=-N}^{N}\frac{1}{(u+k)^2}$$

where $-u$ is a pole of order 2. and $n \in \Bbb{Z}$ and $|n| \leq N$ are poles of order 1. because $$\left|\int \frac{\pi \cot \pi z}{(u+z)^2}\text{dz}\right|\leq \int_0^{2\pi}\pi\left|1+\frac{2}{e^{i2\pi z}-1}\right|\frac{1}{|((N+1/2)e^{i\varphi}+u)^2|}\text{d}\varphi \rightarrow 0$$ as $N \rightarrow +\infty$

Dose there exists some other way to computer it?

share|cite|improve this question
It is a direct application of Fourier transform and Poisson summation formula. See here for a related problem. – Mhenni Benghorbal Feb 26 '13 at 16:21
Sorry, I've a question. Is this convergence uniform ? – Fardad Pouran Dec 29 '15 at 18:56
up vote 8 down vote accepted

Consider the Weierstrass form of the Gamma function:

$$\frac{1}{\Gamma (x)}=xe^{\gamma x}\prod_{n\geq 1}\left(1+\frac{x}{n}\right)e^{-\frac{x}{n}}$$ Then:

$$\ln\Gamma (x)=-\ln x-\gamma x+\sum_{n\geq 1}\left[\frac{x}{n}-\ln \left(1+\frac{x}{n}\right)\right]$$

$$\frac{d}{dx}\ln\Gamma(x)=-\frac{1}{x}-\gamma+\sum_{n\geq 1}\left(\frac{1}{n}-\frac{1}{n+x}\right)$$

$$\begin{align*}\frac{d^2}{dx^2}\ln\Gamma (x)&=\frac{1}{x^2}+\sum_{n\geq 1}\frac{1}{(n+x)^2}\\[7pt]&=\sum_{n\geq 0}\frac{1}{(n+x)^2}\quad(1)\end{align*}$$

Note that by letting $x\to 1-x$ and $n\to-n-1$ we also have:

$$\frac{d^2}{dx^2}\ln\Gamma(1-x)=\sum_{n\leq -1}\frac{1}{(n+x)^2}\quad(2)$$

Combine $(1)$ and $(2):$

$$ \begin{align*}\sum_{\mathbb{Z}}\frac{1}{(n+x)^2}&=\frac{d^2}{dx^2}\ln\Gamma(1-x)+\frac{d^2}{dx^2}\ln\Gamma(x)\\&=\frac{d^2}{dx^2}\ln \Gamma(x)\Gamma(1-x)\\[7pt]&=\frac{d^2}{dx^2}\ln\pi\csc \pi x\quad(x\not\in\mathbb{Z})\\[7pt]&=\pi^2\csc^2\pi x\end{align*}$$

share|cite|improve this answer
It is a very nice proof and I like it. This identity has lots of applications. – xpaul Apr 25 '13 at 20:48

Let $f(z) =\displaystyle \sum_{n\in \mathbb{Z}} \frac{1}{(n+z)^2}$ and $g(z) = \dfrac{\pi^2}{\sin^2(\pi z)}.$ Both functions are $1$ periodic, so it suffices to understand their behavior in the strip $\mathfrak{R}(z)\in [0,1).$ It is straightforward to verify that both these functions have a singularity only at the origin and are bounded on the strip excluding any neighborhood of the origin.

Now look near $z=0.$ We have $f(z) =\displaystyle \frac{1}{z^2} + \sum_{n\in\mathbb{Z}\setminus\{0\}} \frac{1}{(z+n)^2}.$ By the power series expansion of sine and geometric series, we have $$ g(z) = \frac{\pi^2}{\pi^2 z^2 - (\pi^4/3)z^4 + \mathcal{O}(z^6) } = \frac{1}{z^2} \left( \frac{1}{1- (\pi^4/3) z^2 + \mathcal{O}(z^4) } \right) = \frac{1}{z^2} \left( 1 + (\pi^2/3) z^2 + \mathcal{O}(z^4) \right) = \frac{1}{z^2} +\pi^2/3 + \mathcal{O}(z^2).$$

So in fact $f(z)-g(z)$ is bounded and analytic on the strip, and hence the whole complex plane. By Liouville's theorem, $f(z)-g(z) = f(0)-g(0)=0$ so $f(z) = g(z).$

share|cite|improve this answer
Yes, this is the classical proof from Ahlfors. – Bombyx mori May 29 '14 at 7:46

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.