1
$\begingroup$

Let $p \in \mathbb{N^*}, x \in \mathbb{R}$, someone I know was trying to compute the following sum:

$$\sum_{k=0}^{2p-1}\left( \cos\left(x+\frac{k \pi}{2p}\right)\right)^{2p}$$

It seems that the result is $\frac{2p}{4^p} {2p \choose p}$ (which is notably independent of $x$) but the proof I encountered (by using Euler's formula and expanding with Newton binom) does not seem very natural to me. I wonder if someone here has an elegant way of computing this sum and if it can be generalized to other sums with $\sin$ for instance or other powers.

$\endgroup$
2
  • 2
    $\begingroup$ The $\frac1{4^p}\binom{2p}p$ comes pretty directly from the binomial theorem applied to the Euler formula, and the other terms cancel out because the center of gravity of a regular polygon is the center of the circle. It seems difficult to prove otherwise. Maybe you can prove it is constant by showing the derivative is zero, but that doesn't look easy. $\endgroup$ Commented Sep 16, 2022 at 22:39
  • $\begingroup$ @ThomasAndrews can you elaborate on the geometrical interprétation ? $\endgroup$
    – Lelouch
    Commented Sep 17, 2022 at 14:39

2 Answers 2

3
$\begingroup$

I thought I would try to derive your result. After a number of mistakes, here it is.

I agree - a simpler proof would be nice.

Using $\cos(x) =\frac12(e^{ix}+e^{-ix}) $,

$\begin{array}\\ s_p(x) &=\sum_{k=0}^{2p-1}\left( \cos\left(x+\frac{k \pi}{2p}\right)\right)^{2p}\\ &=\dfrac1{2^{2p}}\sum_{k=0}^{2p-1}\left( e^{i\left(x+\frac{k \pi}{2p}\right)}+e^{-i\left(x+\frac{k \pi}{2p}\right)}\right)^{2p}\\ &=\dfrac1{4^{p}}\sum_{k=0}^{2p-1}\sum_{j=0}^{2p}\binom{2p}{j}\left( e^{ij\left(x+\frac{k \pi}{2p}\right)}e^{-i(2p-j)\left(x+\frac{k \pi}{2p}\right)}\right)\\ &=\dfrac1{4^{p}}\sum_{k=0}^{2p-1}\sum_{j=0}^{2p}\binom{2p}{j}\left( e^{ij\left(x+\frac{k \pi}{2p}\right)-i(2p-j)\left(x+\frac{k \pi}{2p}\right)}\right)\\ &=\dfrac1{4^{p}}\sum_{k=0}^{2p-1}\sum_{j=0}^{2p}\binom{2p}{j}\left( e^{2ij\left(x+\frac{k \pi}{2p}\right)-i2p\left(x+\frac{k \pi}{2p}\right)}\right)\\ &=\dfrac1{4^{p}}e^{-2ipx}\sum_{k=0}^{2p-1}\sum_{j=0}^{2p}\binom{2p}{j}\left( e^{2ij\left(x+\frac{k \pi}{2p}\right)-ik\pi}\right)\\ &=\dfrac1{4^{p}}e^{-2ipx}\sum_{j=0}^{2p}\sum_{k=0}^{2p-1}\binom{2p}{j}\left( e^{2ij\left(x+\frac{k \pi}{2p}\right)-ik\pi}\right)\\ &=\dfrac1{4^{p}}e^{-2ipx}\sum_{j=0}^{2p}e^{2ijx}\binom{2p}{j}\sum_{k=0}^{2p-1}\left( e^{2ij\left(\frac{k \pi}{2p}\right)-ik\pi}\right)\\ &=\dfrac1{4^{p}}e^{-2ipx}\sum_{j=0}^{2p}e^{2ijx}\binom{2p}{j}\sum_{k=0}^{2p-1}\left( e^{\left(\frac{ijk \pi}{p}\right)-ik\pi}\right)\\ &=\dfrac1{4^{p}}e^{-2ipx}\sum_{j=0}^{2p}e^{2ijx}\binom{2p}{j}\sum_{k=0}^{2p-1}\left( e^{k\pi i\left(\frac{j}{p}-1\right)}\right)\\ &=\dfrac1{4^{p}}e^{-2ipx}\left(2pe^{2ipx}\binom{2p}{p}+\sum_{j=0,j\ne p}^{2p}e^{2ijx}\binom{2p}{j}\dfrac{1-e^{2p\pi i\left(\frac{j}{p}-1\right)}}{1-e^{\pi i\left(\frac{j}{p}-2\right)}}\right)\\ &=\dfrac{1}{4^{p}}e^{-2ipx}\left(2pe^{2ipx}\binom{2p}{p}+\sum_{j=0,j\ne p}^{2p}e^{2ijx}\binom{2p}{j}\dfrac{1-e^{2\pi i\left(j-p\right)}}{1-e^{\pi i\left(\frac{j}{p}-2\right)}}\right)\\ &=\dfrac{2p}{4^p}\binom{2p}{p}\\ \end{array} $

$\endgroup$
2
  • 1
    $\begingroup$ It's easier if you write $w=e^{ix}, z=e^{i\pi/(2p)}.$ Writing all those exponents over and over again is the main source of noise. $\endgroup$ Commented Sep 17, 2022 at 15:57
  • $\begingroup$ Good idea, though since I entered it directly into Macdown, it was mainly copying and pasting. $\endgroup$ Commented Sep 18, 2022 at 3:19
0
$\begingroup$

This is not an answer.

What seems to happen is that the result of the summation includes at the same time powers of sines and cosines which combine and finally simplify because $\sin^2(x)+\cos^2(x)=1$ $\large (!!)$

Replacing the cosines by the sines leads to the same beautiful result $\frac{2p}{4^p} {2p \choose p}$

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .