What is the closed form for $\sum_{n=-\infty}^{\infty}\frac{1}{(n-a)^2+b^2}$? We can use Fourier series of $e^{-bx}$ ($|x|<\pi$) to evaluate $\sum_{n=-\infty}^{\infty}\frac{1}{n^2+b^2}$. But this one, it seems to me it is tough to get the closed form.

  • $\begingroup$ Clearly, it's period of period one as a function of $a$. $\endgroup$ – Thomas Andrews May 9 '13 at 14:45
  • $\begingroup$ Yes, I realized that a bit too late. $\endgroup$ – leonbloy May 9 '13 at 14:48
  • 1
    $\begingroup$ A bit tough but it can be done with residue theorem. Check math.stackexchange.com/questions/384780/… $\endgroup$ – Philippe Malot May 9 '13 at 15:06
  • $\begingroup$ Thank you for telling this. I think it can be solved by using residue theorem. $\endgroup$ – xpaul May 9 '13 at 15:08
  • 2
    $\begingroup$ I think Poisson's summation formula might help. $\endgroup$ – lee May 9 '13 at 15:59

In my opinion, this is a rather nice example of application of the Poisson summation formula: $$ \sum_{n\in\mathbb{Z}}f(n)=\sum_{k\in\mathbb{Z}}\hat{f}(k),\qquad \hat{f}(\nu)=\int_{-\infty}^{\infty}f(x)e^{-2\pi i \nu x}dx.$$ Namely, setting $f(x)=\frac{1}{(x-a)^2+b^2}$, we find \begin{align} \hat{f}(\nu)=\int_{-\infty}^{\infty}\frac{e^{-2\pi i \nu x}dx}{(x-a)^2+b^2}=e^{-2\pi i \nu a}\int_{-\infty}^{\infty}\frac{e^{-2\pi i \nu x}dx}{x^2+b^2}= \frac{\pi}{b}e^{-2\pi i \nu a-2\pi |\nu| b}, \end{align} where we have assumed that $b>0$ and calculated the last integral using residues. Therefore, the sum we are trying to calculate reduces to geometric series: $$\sum_{n\in\mathbb{Z}}\frac{1}{(n-a)^2+b^2}=\frac{\pi}{b}\sum_{k\in\mathbb{Z}}e^{-2\pi i k a-2\pi |k| b}=\frac{\pi\sinh2\pi b}{b\left(\cosh2\pi b-\cos2\pi a\right)}.$$

  • 1
    $\begingroup$ This is a very nice solution by using Poisson Summation Formula. I like it. Thank you O.L. $\endgroup$ – xpaul May 9 '13 at 16:51
  • $\begingroup$ Is that last term in the denom indeed a cosine or also a hyperbolic cosine? $\endgroup$ – imranfat May 9 '13 at 18:27
  • $\begingroup$ a usual cosine. This is basically because $2\pi a$ appears in the exponential of geometric series with a factor of $i$ and $2\pi b$ without it. $\endgroup$ – Start wearing purple May 9 '13 at 18:33
  • $\begingroup$ Allright then... $\endgroup$ – imranfat May 9 '13 at 18:54
  • $\begingroup$ A related problem. $\endgroup$ – Mhenni Benghorbal May 21 '13 at 17:43

This one can also be done with the standard technique of using the $\pi \cot(\pi z)$ multiplier, integrating $$f(z) = \frac{\pi \cot(\pi z)}{(z-a)^2+b^2}$$ along a circle using the same technique as here.

We integrate $f(z)$ along a circle of radius $R$ with $R$ going to infinity and the integral disappears in the limit so that the residues sum to zero (actually a square with vertices $$(\pm(N+1/2),\pm(N+1/2))$$ $N$ a positive integer is easier to handle computationally). The poles of $f(z)$ other than at the integers are at $$z_{0,1} = a\pm ib$$ and the residues are $$\operatorname{Res}(f(z); z=z_0) = \left.\frac{\pi \cot(\pi z)}{2(z-a)}\right|_{z=z_0} = \frac{\pi\cot(\pi(a+bi))}{2bi} = \frac{\pi}{2bi} i \frac{e^{i\pi(a+bi)}+ e^{-i\pi(a+bi)}}{e^{i\pi(a+bi)}- e^{-i\pi(a+bi)}}$$ and $$\operatorname{Res}(f(z); z=z_1) = \left.\frac{\pi \cot(\pi z)}{2(z-a)}\right|_{z=z_1} = \frac{\pi\cot(\pi(a-bi))}{-2bi} = \frac{\pi}{-2bi} i \frac{e^{i\pi(a-bi)}+ e^{-i\pi(a-bi)}}{e^{i\pi(a-bi)}- e^{-i\pi(a-bi)}}.$$ Now put $x=e^{i\pi a}$ and $y=e^{-\pi b}.$ The first residue becomes $$\operatorname{Res}(f(z); z=z_0) = \frac{\pi}{2b}\frac{xy+1/x/y}{xy-1/x/y}$$ and the second residue is $$\operatorname{Res}(f(z); z=z_1) = -\frac{\pi}{2b}\frac{x/y+y/x}{x/y-y/x}.$$

Adding the two contributions and simplifying with Euler's formula yields $$-\frac{\pi x^2 (y^4-1)}{b (x^2y^2-1)(x^2-y^2)} = -\frac{\pi (y^2-1/y^2)}{b (x^2-1/y^2)(1-y^2/x^2)} \\= -\frac{\pi (y^2-1/y^2)}{b (x^2+1/x^2-y^2-1/y^2)} = \frac{\pi}{b} \frac{\sinh(2\pi b)}{\cos(2\pi a)-\cosh(2\pi b)}.$$

Now with $S$ being our sum we have by the Cauchy Residue Theorem that $$ S + \frac{\pi}{b} \frac{\sinh(2\pi b)}{\cos(2\pi a)-\cosh(2\pi b)} = 0$$ so that finally $$ S = \frac{\pi}{b} \frac{\sinh(2\pi b)}{\cosh(2\pi b)-\cos(2\pi a)}.$$

This MSE link contains a similar computation, this one including the computation of the relevant bounds.

  • $\begingroup$ I like this solution too. $\endgroup$ – xpaul May 9 '13 at 21:09

$\newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\dsc}[1]{\displaystyle{\color{#c00000}{#1}}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\Li}[1]{\,{\rm Li}_{#1}} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large A}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert}$ \begin{align}&\color{#66f}{\large% \sum_{n\ =\ -\infty}^{\infty}{1 \over \pars{n - a}^{2} + b^{2}}} =\sum_{n\ =\ -\infty}^{-1}{1 \over \pars{n - a}^{2} + b^{2}} +{1 \over a^{2} + b^{2}} +\sum_{n\ =\ 1}^{\infty}{1 \over \pars{n - a}^{2} + b^{2}} \\[5mm]&=\sum_{n\ =\ 1}^{\infty}{1 \over \pars{n + a}^{2} + b^{2}} +{1 \over a^{2} + b^{2}} +\sum_{n\ =\ 1}^{\infty}{1 \over \pars{n - a}^{2} + b^{2}} \\[5mm]&=-\,{1 \over a^{2} + b^{2}} +\bracks{\color{#c00000}{% \sum_{n\ =\ 0}^{\infty}{1 \over \pars{n - a}^{2} + b^{2}}} + \pars{a \to -a}} \end{align}

\begin{align}&\color{#c00000}{% \sum_{n\ =\ 0}^{\infty}{1 \over \pars{n - a}^{2} + b^{2}}} =\sum_{n\ =\ 0}^{\infty} {1 \over \bracks{n - \pars{a + b\ic}}\bracks{n - \pars{a - b\ic}}} \\[5mm]&={\Psi\pars{a + b\ic} - \Psi\pars{a - b\ic}\over \pars{a + b\ic} - \pars{a - b\ic}} ={2\ic\,\Im\Psi\pars{a + b\ic} \over 2b\ic}={\Im\Psi\pars{a + b\ic} \over b} \end{align}

\begin{align}&\color{#66f}{\large% \sum_{n\ =\ -\infty}^{\infty}{1 \over \pars{n - a}^{2} + b^{2}}} =\color{#66f}{\large -\,{1 \over a^{2} + b^{2}} + {\Im\Psi\pars{a + b\ic} + \Im\Psi\pars{-a + b\ic} \over b}} \end{align}

$\ds{\Psi\pars{z}}$ is the Digamma Function .



  • Differentiate the natural logarithm of Euler's infinite product formula for the sine function. The logarithm of a product is the sum of logarithms, and the derivative of a sum is the sum $~$ of derivatives.

  • Then, returning to your series, factor the denominator, using complex numbers if necessary, and expand the summand with the help of partial fractions. Then break up the sum into two series, both of which will be expressible in closed forms using the previous result.

  • Employ Euler's formula, linking trigonometric functions of complex argument to hyperbolic ones, in order to furbish the final result.

Euler : $\quad\displaystyle\sin\pi x~=~\pi x~\prod_{\large n\in\mathbb Z}^{\large n\neq0}\bigg(1-\frac xn\bigg)\quad=>\quad\ln\sin\pi x~=~\ln\pi x~+~\sum_{\large n\in\mathbb Z}^{\large n\neq0}\ln\bigg(1-\frac xn\bigg)$.

Differentiating, we have $~\displaystyle\pi\cot\pi x~=~\sum_{\large n\in\mathbb Z}~\frac1{x-n}.~$ At the same time, $~(n-a)^2+b^2~$ can be

factored into $\Big[(n-a)-{\bf i}b\Big]\Big[(n-a)+{\bf i}b\Big]=\Big[n-(a+{\bf i}b)\Big]\Big[n-(a-{\bf i}b)\Big]=(n-z)(n-\bar z)$,

allowing us after partial fraction decomposition to write the sum as $~\displaystyle\frac{\bf i}{2b}~\sum_{\large n\in\mathbb Z}\bigg[\frac1{n-\bar z}-\frac1{n-z}\bigg]$

which then becomes $\dfrac{{\bf i}\pi}{2b}\Big(\cot\pi z-\cot\pi\bar z\Big).~$ Now all that's left to do is employing the formula

for $\cot\Big(A\pm B\Big)~=~\dfrac{\cot A\cot B\mp1}{\cot B\pm\cot A},~$ where $A=a\pi$ and $B={\bf i}b\pi,~$ and $\cot{\bf i}x=-{\bf i}\coth x$.


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.