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

Can \begin{equation} \sum_{k\geq 0}\frac{\left( -1\right) ^{k}\left( 2k+1\right) }{\left( 2k+1\right) ^{2}+a^{2}}, \end{equation} be summed explicitly, where $a$ is a constant real number? If $a=0,$ this sum becomes \begin{equation} \sum_{k\geq 0}\frac{\left( -1\right) ^{k}}{2k+1}=\frac{\pi }{4}. \end{equation} What about for $a\neq 0$. I tried this method \begin{eqnarray*} \sum_{k\geq 0}\frac{\left( -1\right) ^{k}\left( 2k+1\right) }{\left( 2k+1\right) ^{2}+a^{2}} &=&\frac{1}{2}\sum_{k\geq 0}\left( -1\right) ^{k}% \left[ \frac{1}{2k+1+ja}+\frac{1}{2k+1-ja}\right] \\ &=&\frac{1}{2}\sum_{k\geq 0}\frac{\left( -1\right) ^{k}}{2k+1+ja}+\frac{1}{2}% \sum_{k\geq 0}\frac{\left( -1\right) ^{k}}{2k+1-ja} \\ &=&\frac{1}{2}\sum_{k\geq 0}\left( -1\right) ^{k}\int_{0}^{1}x^{2k+ja}dx+% \frac{1}{2}\sum_{k\geq 0}\left( -1\right) ^{k}\int_{0}^{1}x^{2k-ja}dx \\ &=&\frac{1}{2}\int_{0}^{1}\left[ \sum_{k\geq 0}\left( -1\right) ^{k}x^{2k+ja}% \right] dx+\frac{1}{2}\int_{0}^{1}\left[ \sum_{k\geq 0}\left( -1\right) ^{k}x^{2k-ja}\right] dx \\ &=&\frac{1}{2}\int_{0}^{1}\frac{x^{ja}}{1+x^{2}}dx+\frac{1}{2}\int_{0}^{1}% \frac{x^{-ja}}{1+x^{2}}dx \\ &=&\frac{1}{2}\int_{0}^{1}\frac{x^{ja}+x^{-ja}}{1+x^{2}}dx \end{eqnarray*} but I was stucked at the last equations. Can any one give me some hint or tell me that the analytic expression doesn't exist. Thanks very much!

share|cite|improve this question
Note for $x>0$: $x^{ia}+x^{-ia}=e^{ia\log(x)}+e^{-ia\log(x)}=2\cos(a\log(x))$. – Your Ad Here Oct 31 '12 at 20:57
up vote 5 down vote accepted

The answer is : $$S(a)=\sum_{k\geq 0}\frac{\left( -1\right) ^{k}\left( 2k+1\right) }{\left( 2k+1\right) ^{2}+a^{2}}=\frac {\pi}{4\,\cosh(\pi a/2)}$$

Let's start nearly as you did : \begin{eqnarray*} S(a)=\sum_{k\geq 0}\frac{\left( -1\right) ^{k}\left( 2k+1\right) }{\left( 2k+1\right) ^{2}+a^{2}} &=&\frac 12\sum_{k\geq 0}\left( -1\right) ^{k}% \left[ \frac 1{2k+1+ia}-\frac 1{-2k-1+ia}\right] \\ &=&\frac 12% \sum_{k=-\infty}^\infty \frac{\left( -1\right) ^{k}}{2k+1+ia} \\ &=&\frac 14% \sum_{k=-\infty}^\infty \frac{\left( -1\right) ^{k}}{k+(1+ia)/2} \\ \end{eqnarray*}

But the last series was evaluated earlier here using different methods with the result (in our case) : $$S(a)=\frac {\pi}{4\sin\left(\pi (1+ia)/2\right)}=\frac {\pi}{4\cos\left(\pi a i/2\right)}=\frac {\pi}{4\cosh\left(\pi a/2\right)}$$

Since your method is fine let's try to finish it your way :

$$ \begin{align} S(a)&=\frac 12\int_{0}^{1}\frac{x^{ia}+x^{-ia}}{1+x^{2}}dx\\ &=\int_{0}^{1}\frac{\cos(a\ln(x))}{1+x^{2}}dx\quad\text{using the hint in the comment}\\ &=\int_{-\infty}^0\frac{\cos(a\,t)\,e^t}{1+e^{2t}}dt\quad\text{if}\ \ x:=e^t\\ &=\frac 12\int_{-\infty}^0\frac{\cos(a\,t)}{\cosh(t)}dt\\ &=\frac 12\int_0^\infty \frac{\cos(a\,t)}{\cosh(t)}dt\\ \end{align} $$ Let's evaluate $\displaystyle I(a)=\int_{-R}^R \frac{e^{i a\,z}}{\cosh(z)}dz\ $ using a rectangular contour $[-R,+R]\times [0,\pi i]\ $ (we will have $R\to\infty$ and suppose $|a|<1$)

$$ I(a)=\int_{-R}^R \frac{e^{i a\,z}}{\cosh(z)}dz+\int_R^{R+\pi i} \frac{e^{i a\,z}}{\cosh(z)}dz+\int_{R+\pi i}^{-R+\pi i}\frac{e^{i a\,z}}{\cosh(z)}dz+\int_{-R+\pi i}^{-R} \frac{e^{i a\,z}}{\cosh(z)}dz $$ putting the third integral in second position : $$ I(a)=\int_{-R}^R \frac{e^{i a\,t}}{\cosh(t)}dt+\int_{R}^{-R}\frac{e^{i a(t+\pi i)}}{\cosh(t+\pi i)}dt+\int_R^{R+\pi i} \frac{e^{i a\,z}}{\cosh(z)}dz+\int_{-R+\pi i}^{-R} \frac{e^{i a\,z}}{\cosh(z)}dz $$ I'll let you show that the two integrals at the right disappear as $R\to\infty$ (for $|a|<1$) and use $\cosh(t+\pi i)=-\cosh(t)$ to get : $$I(a)=\int_{-\infty}^\infty \frac{e^{i a\,t}}{\cosh(t)}dt+\int_{-\infty}^\infty\frac{e^{i a(t+\pi i)}}{\cosh(t)}dt=\left(1+e^{-a\pi}\right)\int_{-\infty}^\infty \frac{e^{i a\,t}}{\cosh(t)}dt$$

Since there is only one pole at $\ t=\frac {\pi}2 i\ $ we may use the residu there to get : $$\frac{2\pi i\, e^{i a\,\pi i/2}/i}{\left(1+e^{-\pi a}\right)}=\frac{2\pi\, e^{-\pi a/2}}{\left(1+e^{-\pi a}\right)}=\frac{\pi}{\cosh(\pi a/2)}=\int_{-\infty}^\infty \frac{e^{i a\,t}}{\cosh(t)}dt$$ and conclude that indeed :

$$\frac 12\int_0^\infty \frac{\cos(a\,t)}{\cosh(t)}dt=\frac{\pi}{4\cosh(\pi a/2)}$$

Perhaps not as simple as you hoped...

share|cite|improve this answer
Can you tell me how to derive the result? – Xiangyu Meng Oct 31 '12 at 23:19
@XiangyuMeng: yes I am redacting a more complete answer. – Raymond Manzoni Oct 31 '12 at 23:21
I checked both of your answers. It requires that |a|<1. Does the result still hold for |a|>1? – Xiangyu Meng Nov 1 '12 at 22:35
@XiangyuMeng: for the last derivation I supposed $|a|<1$ for convenience. Supposing that $|\Im{a}|<1$ should be enough (note that $a=i$ would give $0$ in your series' denominator) since the real part contributes only a $e^{i\Re{a}}$ part in the two neglected integrals at the right. Anyway you may use analytic extension to extend the rather smooth result as wished. – Raymond Manzoni Nov 1 '12 at 23:07
I assume that $a$ is a real number, thus the |ℑa|<1 would be satisfied. Thanks very much for your answer. – Xiangyu Meng Nov 1 '12 at 23:14

This sum can be evaluated by the same trick as presented here:

In order to capture the convergence behavior of the series group consecutive terms to obtain absolute convergence, writing $$ S(a) = \sum_{m\ge 0} \frac{4m+1}{(4m+1)^2+a^2} - \sum_{m\ge 0} \frac{4m+3}{(4m+3)^2+a^2} = \sum_{m\ge 0} \frac{(4m+1)((4m+3)^2+a^2)-(4m+3)((4m+1)^2+a^2)} {((4m+3)^2+a^2)((4m+1)^2+a^2)} = \sum_{m\ge 0} \frac{2(4m+1)(4m+3)-2a^2}{((4m+3)^2+a^2)((4m+1)^2+a^2)} $$

Instead of using $f_1(z)$ and $f_2(z)$ from the other post use $$f(z) = \frac{2(4z+1)(4z+3)-2a^2}{((4z+3)^2+a^2)((4z+1)^2+a^2)} \pi \cot(\pi z).$$

The key operation of this technique is to compute the integral of $f(z)$ along a circle of radius $R$ in the complex plane, where $R$ goes to infinity. We certainly have $|\pi\cot(\pi z)|<2\pi$ for $R$ large enough. The core term from the sum is $\theta(R^2/R^4)$ which is $\theta(1/R^2)$ so that the integrals are $\theta(1/R)$ and vanish in the limit. This means that the sum of the residues at the poles of $f(z)$ add up to zero.

It is easily verified that the poles at the integers produce the terms of the sum twice over. It follows that $$ 2 S(a) + \text{Res}_{z=\frac{3}{4}-\frac{1}{4}ia} f(z) + \text{Res}_{z=\frac{3}{4}+\frac{1}{4}ia} f(z) + \text{Res}_{z=\frac{1}{4}-\frac{1}{4}ia} f(z) + \text{Res}_{z=\frac{1}{4}+\frac{1}{4}ia} f(z) = 0$$

The residues are easily computed as the poles are all simple. This finally yields $$ S(a) = \frac{1}{4} \frac{\pi}{\cosh \left(\frac{\pi a}{2}\right)}.$$

share|cite|improve this answer

The series is summable and has the closed form formula

$$-\frac{1}{4}\,{\frac {\pi }{\sin \left( \frac{1}{2}\pi \, \left( 3 + ia \right ) \right) }} = \frac{1}{4}{\frac {\pi }{\cosh \left( \frac{\pi \,a}{2} \right) }} .$$

share|cite|improve this answer
Can you tell me how to derive the result? – Xiangyu Meng Oct 31 '12 at 23:19
@did: I hope you can see it now! – Mhenni Benghorbal Nov 2 '12 at 19:02
@did:Lift up your downvote. – Mhenni Benghorbal Nov 2 '12 at 19:03
@did:did: I do not know what to say, but if you do not know something, do not talk about it – Mhenni Benghorbal Nov 2 '12 at 20:40

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.