On the Wolfram Research Reference page for the cotangent function (https://functions.wolfram.com/ElementaryFunctions/Cot/23/01/), I saw the following partial sum formula

$$\sum_{k=0}^{n-1}(-1)^k\cot\Big(\frac{\pi}{4n}(2k+1)\Big)=n.$$ I was unable to find a reference for it but eventually proved it as described below.

My Question

  1. A reference for the equation above?
  2. Naturally, I wondered what was known about the regular sum of the cotangents above, not the alternating sums. I found no related equation but instead was able to empirically derive the following limit

\begin{align*} \lim_{n\to\infty}\Big[\sum_{k=0}^{n-1}\frac{1}{2k+1}-\sum_{k=0}^{n-1}\frac{\pi}{4n}\cot\Big(\frac{\pi}{4n}(2k+1)\Big)\Big]&=\lim_{n\to\infty}\Big[\Big(H_{2n}-\frac{1}{2}H_{n}\Big)-\sum_{k=0}^{n-1}\frac{\pi}{4n}\cot\Big(\frac{\pi}{4n}(2k+1)\Big)\Big]\\ &=\ln(\sqrt{\pi/2}). \end{align*}

Yet, I am stuck on proving this limit and would appreciate any advice on where to begin.

My Proof for the Equation in the Beginning:

Consider the well-known partial fraction expansion for the cotangent function,

$$\pi\cot(\pi x)=\frac{1}{x}+\sum_{k=1}^{\infty}\frac{1}{x-k}+\frac{1}{x+k}$$

excluding integral $x$. Rather unconventionally, let us expand the summation of the RHS and reindex it, getting \begin{align*} \frac{1}{x}+\frac{1}{x-1}+\frac{1}{x+1}+\frac{1}{x-2}+\frac{1}{x+2}+\cdots&=\frac{1}{x}-\frac{1}{1-x}+\frac{1}{x+1}-\frac{1}{2-x}+\frac{1}{x+2}-\cdots\\ &=\sum_{k=0}^{\infty}\frac{1}{k+x}-\frac{1}{k+1-x}. \end{align*} By letting $x=b/a$ for positive integers $a,b$ such that $a>b\geq1,$ we get


Consider the equation for $a=4n$ and $b$ from $1$ to $2n-1$. Then, $$\frac{\pi}{4n}\cot\Big(\pi\cdot\frac{1}{4n}\Big)=\sum_{k=0}^{\infty}\frac{1}{(4n)k+1}-\frac{1}{(4n)k+4n-1},$$ $$\frac{\pi}{4n}\cot\Big(\pi\cdot\frac{3}{4n}\Big)=\sum_{k=0}^{\infty}\frac{1}{(4n)k+3}-\frac{1}{(4n)k+4n-3},$$ $$\cdots$$ $$\frac{\pi}{4n}\cot\Big(\pi\cdot\frac{2n-1}{4n}\Big)=\sum_{k=0}^{\infty}\frac{1}{(4n)k+2n-1}-\frac{1}{(4n)k+2n+1}.$$

By summing up the equations in an alternating fashion (adding the first, subtracting the second, and so on,) the summed $LHS$ equals $$\frac{\pi}{4n}\sum_{k=0}^{n-1}(-1)^k\cot\Big(\frac{\pi}{4n}(2k+1)\Big)$$ while the summed RHS is simply the alternating series of the reciprocals of odd numbers, which is well-known by Leibniz to be $\frac{\pi}{4}.$ Hence, we get the initial formula.

  • 2
    $\begingroup$ These kind of identities are well-known (folklore). For example, one of the homework problems in my introductory complex analysis course is the following. Let $n=2m+1$ be an odd positive integer. Then $\sum_{k=-m}^m\tan(z+k\pi/n)=n\tan(nz)$. (Specialize this to $z=\pi/(4n)$.) $\endgroup$
    – GH from MO
    Aug 10 '20 at 13:49
  • 2
    $\begingroup$ I googled "alternating sum odd cotangent", and found this, by Bruce Berndt and Boon Pin Yeap, which has a good set of references for identities like this: sciencedirect.com/science/article/pii/S0196885802000209 $\endgroup$
    – Matt F.
    Aug 10 '20 at 14:01

The expression under the limit sign in question is just a Riemann sum for $$\int_0^{\pi/2} \frac12\,\Big(\frac1x-\cot x\Big)\,dx=\frac12\,\ln\frac\pi2,$$ which therefore is the value of the limit.

  • $\begingroup$ Just to make sure, the definite integral on the left gives $-\ln(\sin(1))$? $\endgroup$
    – bryanjaeho
    Aug 10 '20 at 15:41
  • 1
    $\begingroup$ I believe the integral should be over $\frac{1}{2}\Big(\frac{1}{x}-\frac{\pi}{2}\cdot \cot(\pi/2\cdot x)\Big)$ ranging from $0$ to $1$. Thank you for the approach! $\endgroup$
    – bryanjaeho
    Aug 10 '20 at 16:00
  • 1
    $\begingroup$ @bryanjaeho : Oops! The integral is of course from$0$ to $\pi/2$, not to $1$. This typo is now corrected $\endgroup$ Aug 10 '20 at 16:11

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.