We want to evaluate the sums $$S_{n,p}=\sum_{k=1}^{n}\sin^{2p}\left({k\pi\over 2n+1}\right)$$ So we try many attempts by experimenting with the sum calculator which leads us to the result $$S_{n,p}={{2p \choose p}(2n+1)\over 2^{2p+1}}$$ How do we proof that the above result is correct or not?
-
Is $p$ taken to be a natural number? – Michael L. Sep 12 '17 at 20:44
-
Test with $n:=1$ : Is this correct ?!? – user90369 Sep 12 '17 at 20:44
-
1$n=1$ and $p=3$ does not work – GuPe Sep 12 '17 at 20:46
-
@GuachoPerez : That's what I meant but I hoped the OP will test it himself. :) – user90369 Sep 12 '17 at 20:49
-
It seems like it might work for $n > 1$, though. – Michael L. Sep 12 '17 at 20:53
-
Mathematica shows it works only for $p=2$ and $n=1$, i checked it for $p<10$. – Nosrati Sep 12 '17 at 20:59
1 Answers
Assuming that $p$ is a positive integer and letting $\omega = e^{\pi i/(2n+1)}$, we have
\begin{align*} \sin^{2p}\left( \frac{k\pi}{2n+1} \right) &= \frac{1}{2^{2p}} \sum_{j=0}^{2p} \binom{2p}{j} (-1)^{p-j} \omega^{2k(p-j)} \\ &= \frac{1}{2^{2p+1}} \sum_{j=0}^{2p} \binom{2p}{j} (-1)^{p-j} \left( \omega^{2k(p-j)} + \omega^{-2k(p-j)} \right). \end{align*}
Now utilizing the Dirichlet kernel formula, we obtain
$$ \sum_{k=1}^{n} \left( \omega^{2k(p-j)} + \omega^{-2k(p-j)} \right) = -1 + \sum_{k=-n}^{n} \omega^{2k(p-j)} = \begin{cases} 2n, & 2n+1 \mid p-j \\ -1, & 2n+1 \nmid p-j \end{cases}.$$
So it follows that
\begin{align*} \sum_{k=1}^{n} \sin^{2p}\left( \frac{k\pi}{2n+1} \right) &= \frac{1}{2^{2p+1}} \sum_{j=0}^{2p} \binom{2p}{j} (-1)^{p-j} \left( (2n+1)\mathbf{1}_{\{ 2n+1 \mid p-j \}} - 1 \right) \\ &= \frac{2n+1}{2^{2p+1}} \sum_{l} (-1)^{l} \binom{2p}{p+(2n+1)l}, \end{align*}
where the sum in the last line runs over all integers $l$ such that $|l| \leq \frac{p}{2n+1}$. In particular, if $p \leq 2n$ then only $l = 0$ contributes to the sum and we retrieve OP's formula.

- 167,468
-
Also check my previous computation: https://math.stackexchange.com/questions/1932029/how-to-evaluate-sum-limits-k-0n-1-sint-pi-k-2n/1935309#1935309 – Sangchul Lee Sep 12 '17 at 21:31