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 want to show that $\sum_{|j| < n} (n-|j|) \exp(ij\lambda)= \dfrac{\sin^2(\frac 1 2 n\lambda)}{\sin^2(\frac 1 2 \lambda)}$.

I know from Proving $\sum\limits_{k=0}^{n}\cos(kx)=\frac{1}{2}+\frac{\sin(\frac{2n+1}{2}x)}{2\sin(x/2)}$ \begin{equation} (1)\sum_{j=1}^{n-1} cos(j\lambda) = -\frac 1 2 + \frac{\sin(\frac{2n-1} 2 \lambda)}{2\sin(\frac \lambda 2)}. \end{equation}

and from the Hint in the first answer in How to show that $\frac{1}{\tan(x/2)}=2 \sum_{j=1}^{\infty}\sin(jx)$ in Cesàro way/sense? $$ (2)\frac d {dx} \sum_{j=1}^{n-1} sin(j\lambda) =\frac d {d\lambda} \frac{\cos({\frac \lambda 2})-\cos(\frac{n+1}2 \lambda)}{2\sin(\frac\lambda 2)} = \frac{n\sin(\frac{n+1} 2 \lambda)\sin(\frac \lambda 2)+\cos(\frac{n\lambda}2) - 1}{4\sin^2(\frac \lambda 2)}. $$

This is what i did so far: \begin{align} \sum_{|j| < n} (n-|j|) \exp(ij\lambda) &= n + 2\sum_{j=1}^{n-1}(n-j) cos(j\lambda) \\ &=n + 2n \sum_{j=1}^{n-1} cos(j\lambda) - \frac d {d\lambda}\sum_{j=1}^{n-1} sin(j\lambda)\\ &= \frac{n\sin(\frac{2n-1} 2 \lambda)\sin(\frac \lambda 2) + \frac 1 2 + \frac 1 2 n \sin(\frac {n+1} 2 \lambda)\sin(\frac \lambda 2) + \frac 1 2 \cos( \frac{n\lambda}2)}{\sin^2(\frac \lambda 2)}. \end{align}

If this and the proposition is true, it should be true that $$ n\sin(\frac{2n-1} 2 \lambda)\sin(\frac \lambda 2) + \frac 1 2 + \frac 1 2 n \sin(\frac {n+1} 2 \lambda)\sin(\frac \lambda 2) + \frac 1 2 \cos( \frac{n\lambda}2) - \sin^2(\frac {n\lambda} 2) = 0. $$

But for $n=2$, i get $$ 3\sin(\frac {3\lambda} 2) \sin(\frac \lambda 2) + \frac 1 2 + \frac 1 2 \cos(\lambda) - \sin^2 \lambda \ne 0. $$ (checked with Wolframalpha)

So there is probably something wrong with my calculations. But i checked everything twice and don't see my mistake. Do you see it?


Trying to prove it as proposed in the second answer.

Let $z:=\exp(i\lambda)$ and $p(z):=\sum_{j=0}^{n-1} z^j$. First i want to show that $p(z)p(z^{-1})=\sum_{|j|<n} (n-|j|) z^j$.

It is \begin{align} \sum_{j=1}^{n-1} z^{j}&= p(z) - 1 = \frac{1-z^n}{1-z} - 1.\\ \sum_{j=1}^{n-1} z^{-j}&= p(z^{-1})-1 = \frac{1-z^{-n}}{1-z^{-1}} - 1.\\ \sum_{j=1}^{n-1} jz^{-j} &= i \frac d {d\lambda} p(z^{-1}) = \frac{-nz^{-n}+(n-1)z^{-n-1}+z^{-1}}{(1-z^{-1})^2}.\\ \sum_{j=1}^{n-1} jz^{j} &= -i \frac d {d\lambda} p(z) = \frac{-nz^{n}+(n-1)z^{n+1}+z}{(1-z)^2}.\\ \end{align} So i have \begin{align} \sum_{|j|<n} (n-|j|) z^j = n + n(\sum_{j=1}^{n-1} z^{j} + \sum_{j=1}^{n-1} z^{-j}) - \sum_{j=1}^{n-1} jz^{j} - \sum_{j=1}^{n-1} jz^{-j}. \end{align} If i now put in the values from above and substract $p(z)p(z^{-1})$ i should get $0$. But Wolfram Alpha doesn't agree, so i am again on the wrong track. What did i wrong?

The last step $p(z)p(z^{-1})= \frac{(z^{n/2}-z^{-n/2})^2}{(z^{1/2}-z^{-1/2})^2}$ was easy to show.


I did it now. It was a lot easier to show directly $p(z)p(z^{-1})=\sum_{|j|<n} (n-|j|) z^j$ without the geometric series partial sums values.

share|cite|improve this question
It's often easier to solve these sorts of problems by setting $z=e^{i\lambda}$ and then simplifying a resulting polynomial expression. – Thomas Andrews Nov 30 '12 at 14:32
You didn't distribute the $2$ when you expanded $2\sum (n-j)\cos j\lambda$ – Thomas Andrews Nov 30 '12 at 15:38
up vote 2 down vote accepted

Letting $\alpha=\frac{\lambda}{2}$ to make the expressions cleaner, we get: $$\begin{align} \sum_{|j| < n} (n-|j|) \exp(ij\lambda) &= n + 2\sum_{j=1}^{n-1}(n-j) \cos j\lambda \\ &=n + 2n \sum_{j=1}^{n-1} \cos j\lambda - 2\frac d {d\lambda}\sum_{j=1}^{n-1} \sin j\lambda\\ &= \frac{n\sin(2n-1)\alpha\sin\alpha + \frac 1 2 - \frac 1 2 n \sin (n+1)\alpha \sin \alpha - \frac 1 2 \cos n\alpha}{\sin^2\alpha}. \end{align}$$ Note the minus signs in the numerator. I'm not sure if that solves the problem - there might be an error elsewhere. Wolfram Alpha says when $n=2$ the numerator is $\sin^2 2\alpha$, but $n=1$ doesn't agree.

The easier way to do this sort of problem is via the comment I gave above, using $z=e^{i\lambda}$.

Let $p(z)=\sum_{j=0}^{n-1} z^j$. Show that:$$p(z)p(z^{-1})=\sum_{|j|<n} (n-|j|) z^j$$

Now, $p(z)=\frac{z^n-1}{z-1}$, and $p(z^{-1})=z^{-(n-1)}p(z)$. Now show: $$p(z)p(z^{-1})= \frac{(z^{n/2}-z^{-n/2})^2}{(z^{1/2}-z^{-1/2})^2}$$

Then let $z=e^{i\lambda}$.

share|cite|improve this answer
I tried to proof it your way, but failed again... – Runje Dec 2 '12 at 14:50
I edited my originial post where i tried to proof it. Or is there an easier way to prove it than i tried it? – Runje Dec 2 '12 at 16:16
Thx Thomas, i did it now your way. – Runje Dec 3 '12 at 14:46

The only thing that I see at first sight is that you didn't distribute properly the minus sign in front of the derivative.

share|cite|improve this answer

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.