How can you prove that:

$$\sum_{k=1}^{n-1}\tan^{2}\frac{k \pi}{2n} = \frac{(n-1)(2n-1)}{3}$$

for every integer $n\geq 1$.

PS: no, it's not a homework... :-)

  • $\begingroup$ What is PS. $\endgroup$ – anonymous Aug 13 '10 at 13:30
  • 2
    $\begingroup$ I think it is still good to either explain the motivation behind the question or where it is from $\endgroup$ – Casebash Aug 13 '10 at 13:34
  • 1
    $\begingroup$ @Chandru: english.stackexchange.com :) (PS: PS means [ post scriptum ](en.wiktionary.org/wiki/post_scriptum) (footnote).) $\endgroup$ – kennytm Aug 13 '10 at 14:25
  • $\begingroup$ I've been playing with trigonometric sums and this one yield an interesting value :) $\endgroup$ – botismarius Aug 13 '10 at 17:10
  • 1
    $\begingroup$ The result $n(2n+1)$ for the corresponding sum with odd denominator $2n+1$ is calculated in this answer. $\endgroup$ – joriki Jul 21 '12 at 17:22

By a well know formula we have $$ \left(\cos{\frac{k\pi}{2n}} + i\sin{\frac{k\pi}{2n}}\right)^{2n}=(-1)^{k} $$ Hence by Binomial theorem we have ($[x]$ is not an integer part, brackets are added for clarity) $$ \sum_{t=0}^{2n}\binom{2n}{t}\left[\cos{\frac{k\pi}{2n}}\right]^t \cdot \left[i\sin{\frac{k\pi}{2n}}\right]^{2n-t}=(-1)^{k} $$ Now we consider only imaginary part of this: $$ \sum_{r=0}^{n-1}\binom{2n}{2r+1}\left[\cos{\frac{k\pi}{2n}}\right]^{2r+1} \cdot \left[i\sin{\frac{k\pi}{2n}}\right]^{2n-2r-1}=0 $$ Divide it by $[\cos{\frac{k\pi}{2n}}]^{2n}$: $$ \sum_{r=0}^{n-1}\binom{2n}{2r+1}\left[i\tan{\frac{k\pi}{2n}}\right]^{2n-2r-1}=0 $$ Now multiply by $i\tan{\frac{k\pi}{2n}}$: $$ \sum_{r=0}^{n-1}\binom{2n}{2r+1}\left[i\tan{\frac{k\pi}{2n}}\right]^{2n-2r}=0 $$ So $[\tan{\frac{k\pi}{2n}}]^2$ are roots of the following polynomial: $$ \sum_{r=0}^{n-1}\binom{2n}{2r+1}\left[-x\right]^{n-r}=0 $$ Hence by Vieta's_formulas sum of it roots is equal to $$ \frac{\binom{2n}{3}}{\binom{2n}{1}} =\frac{(2n-1)(n-1)}{3} $$

  • 4
    $\begingroup$ Awesome proof!!! $\endgroup$ – anonymous Aug 13 '10 at 13:25
  • $\begingroup$ Oh wow, that's neat! $\endgroup$ – Alexander Vlasev Oct 20 '12 at 5:07
  • $\begingroup$ very nice proof!. $\endgroup$ – jdoicj Jun 4 '13 at 5:06
  • $\begingroup$ I added my answer and just read your proof. Nice answer! Euler's Formula, the Binomial Formula, and Vieta's Formula. The only complex analysis is in Euler's Formula. $\endgroup$ – robjohn Aug 22 '13 at 19:59
  • $\begingroup$ @robjohn not even complex analysis is needed because de moivre theorem can be proved by induction. $\endgroup$ – Shubhashish Mar 21 '18 at 13:35

This problem is over 3 years old, but I was just made aware of it, so I thought I would add a new method of attack, similar to this answer.

Here we use the function $\dfrac{2n/z}{z^{2n}-1}$ which has residue $1$ at each $2n^\text{th}$ root of unity and residue $-2n$ at $0$.

Since $\tan^2\left(\dfrac\theta2\right)=-\left(\dfrac{z-1}{z+1}\right)^2$ for $z=e^{i\theta}$, define $$ f(z)=-\left(\frac{z-1}{z+1}\right)^2\frac{2n/z}{z^{2n}-1} $$ The sum of the residues of $f$ is $0$ since $|f(z)|\sim2n/z^{2n+1}$ as $|z|\to\infty$ and so an integral around a large circle vanishes.

On the other hand, the sum of the residues is twice the sum we want plus the residue at $0$ and the residue at $-1$.

$\hspace{32mm}$enter image description here

Thus, $$ \begin{align} \sum_{k=1}^{n-1}\tan^2\left(\frac{k\pi}{2n}\right) &=-\frac12\left(\operatorname*{Res}_{z=0}(f(z))+\operatorname*{Res}_{z=-1}(f(z))\right)\\ &=-\frac12\left(2n-\frac23(2n^2+1)\right)\\ &=\frac{2n^2-3n+1}{3} \end{align} $$

  • 1
    $\begingroup$ Very beautiful!!! $\endgroup$ – user 1591719 Aug 22 '13 at 19:19
  • $\begingroup$ Residue methods are my favourite. $\endgroup$ – ireallydonknow Dec 29 '13 at 19:31

I am doing a similar thing for $\cot$ i hope you can reciprocate it.

We have $$(\cot(\theta)-i)^{n} = \frac{ \cos{n\theta} - \sin{n \theta}}{\sin^{n}(\theta)}$$ Equating the real and imaginary parts on both sides we have $$ \frac{\sin(n\theta)}{\sin{\theta}} = \sum\limits_{s} {n \choose 2s+1} (-1)^{s} \cot^{n-2s-1}(\theta)$$

Now take $2n+1$ instead of $n$ we have, $$\sin{(2n+1)\theta} = \sin^{2n+1}(\theta)P_{n}\cot^{2}(\theta)$$

for $\displaystyle 0 < \theta < \frac{\pi}{2}$ and where $P_{n}$ is the polynomial given by $${ 2n+1 \choose 1}T^{n} - {2n+1 \choose 3}T^{n-1} + \cdots$$

Noting that the zeros of $P_{n}$ are precisely $\displaystyle \frac{r\pi}{2n+1},\ n =1,2,..$, we have the first identity from the sum of the roots formula.

From the inequality $\sin{x} < x < \tan{x}$ we have $$\cot^{2m}(x) < \frac{1}{x^{2m}} < (1 + \cot^{2}(x))^{m}$$ we have $$ \sum\limits_{r=1}^{n} \cot^{2m} \frac{r\pi}{2n+1} < \frac{(2n+1)^2m}{\pi^{2m}} \sum\limits_{r=1}^{n} \frac{1}{r^{2m}} < \sum\limits_{r=1}^{n} \Bigl( 1 + \cot^{2} \frac{r\pi}{2n+1} \Bigr)^{m}$$

Therefore, $$\sum\limits_{r=1}^{n} \Bigl(1+\cot^{2} \frac{r\pi}{2n+1} \Bigr)^{m} = \sum\limits_{r=1}^{n} \cot^{2m} \frac{r\pi}{2n+1} + \mathcal{O}(n^{2m-1})$$

In other words to find $c_{2m}$ where $ \sum\limits_{r=1}^{n} \cot^{2m} \frac{r\pi}{2n+1} = c_{2m}n^{2m} + \mathcal{O}(n^{2m-1})$ it suffices to look at the sum, $$ \cot^{2m} \frac{\pi}{2n+1} + \cot^{2m} \frac{2\pi}{2n+1} + \cdots + cot^{2m} \frac{n\pi}{2n+1}$$

which is the sum $s_{m}$ of $m-th$ powers of the roots of the polynomial $P_{n}$. Then you can complete it using newtons formula for the sum of the roots.


I give a proof of a general case of that type of sum related to binomial coefficients in my post Tan binomial formulas from a set S and its k-subsets


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.