I was motivated by the integral $\displaystyle \int_0^{\pi/12} \ln\tan(x)\,dx$ which was posted here which is equal to $-\frac{2}{3}G$ where is $G$ is Catalan's constant. With this motivation I came up with a result

$$\int_0^{\pi/24}\ln\tan x\,dx=\pi\ln\left(\frac{G\left(\frac{25}{24}\right)G\left(\frac{35}{24}\right)}{G\left(\frac{13}{24}\right)G\left(\frac{23}{24}\right)}\right)-\frac{\pi}{24}\ln\left(\frac{4096\pi^{12}}{\sqrt{2-\sqrt{3}}\left(2+\sqrt{2+\sqrt{3}}\right)^5}\right)$$Notation: $G$ denotes Barnes G-function.

Since Wolfram cannot generate the closed form however, the closed form obtained as per the WA check is found to be correct.

Moreover, I came up with following inequality, the close upper bound of the integral, surprisingly in terms of $ e$ and $G$ which is as follows.

$$\bigg|\int_0^{\pi/24}\ln\tan x \,dx\bigg| < \frac{e^{-1}}{G}\cdots (1)$$ Notation: $e$ denotes Euler's number.

Now I'm curious to know,

$\bullet$ Can we generalize the integral

$$\int_0^{\pi/p} \ln \tan x \,dx\, =\text{?} \; \; p\neq 0 $$

$\bullet$ How to prove $(1)$ using the inequality theorem?

Thank you

  • $\begingroup$ See A class of series acceleration formulae for Catalan's constant, David M. Bradley, arxiv.org/pdf/0706.0356.pdf $\endgroup$ – FDP Jul 23 at 7:54

Just approximations.

Concerning the integral

$$I(t)=\int_0^t \log (\tan (x))\,dx$$ $$I=\tan ^{-1}(\tan (t)) \log (\tan (t))+\frac{i}{2} (\text{Li}_2(i \tan (t))-\text{Li}_2(-i \tan (t)))$$

Since $t$ is supposed to be small, I should use $$I(t)=-t (1-\log (t))+\frac {t^2}9 \sum_{i=1}^\infty \frac{a_i}{b_i}\, t^{2i-1} $$ where the first coefficients are $$\left( \begin{array}{ccc} i & a_i & b_i \\ 1 & 1 & 1 \\ 2 & 7 & 50 \\ 3 & 62 & 2205 \\ 4 & 127 & 18900 \\ 5 & 146 & 81675 \\ 6 & 1414477 & 2766889125 \\ 7 & 32764 & 212837625 \\ 8 & 16931177 & 351486135000 \\ 9 & 11499383114 & 740606614122375 \\ 10 & 91546277357 & 17865510428390625 \end{array} \right)$$

Using $t=\frac \pi {24}$ this gives as a result $-0.39681136139214865267218614585540$ while the exact result should be $-0.39681136139214865267218614585537$


Another possibility is to let $\tan(x)=u$ to make $$I(t)=\int_0^{\tan ^{-1}(t)} \frac{\log (u)}{u^2+1}\,du= \tan ^{-1}(u)\log (u)+\frac{i}{2} (\text{Li}_2(i u)-\text{Li}_2(-i u))$$ and use $$\frac{i}{2} (\text{Li}_2(i u)-\text{Li}_2(-i u))=\sum_{n=0}^\infty \frac{(-1)^{n+1}}{(2 n+1)^2} u^n$$ which has the merit to be alternating.

More compact and exact will be $$\color{red}{I(t)=\tan ^{-1}\left(\tan ^{-1}(t)\right) \log \left(\tan ^{-1}(t)\right)-\frac{1}{4} \tan ^{-1}(t)\, \Phi \left(-\tan ^{-1}(t)^2,2,\frac{1}{2}\right)}$$ where appears the Lerch transcendent function.

| cite | improve this answer | |

Only extended comment.

With CAS help:

$\int_0^{\frac{\pi }{p}} \ln (\tan (x)) \, dx=\tan ^{-1}\left(\tan \left(\frac{\pi }{p}\right)\right) \ln \left(\tan \left(\frac{\pi }{p}\right)\right)-\frac{1}{2} i \text{Li}_2\left(-i \tan \left(\frac{\pi }{p}\right)\right)+\frac{1}{2} i \text{Li}_2\left(i \tan \left(\frac{\pi }{p}\right)\right)$

for: $p > 2$, where: $\text{Li}_2(x)$ is polylogarithm function.

Mathematica code:

Integrate[Log[Tan[x]], {x, 0, Pi/p}] == ArcTan[Tan[\[Pi]/p]] Log[Tan[\[Pi]/p]] - 1/2 I PolyLog[2, -I Tan[\[Pi]/p]] + 1/2 I PolyLog[2, I Tan[\[Pi]/p]]]

| cite | improve this answer | |
  • $\begingroup$ Or simpler: $-\frac{1}{2} i \text{Li}_2\left(-i \tan \left(\frac{\pi }{p}\right)\right)+\frac{1}{2} i \text{Li}_2\left(i \tan \left(\frac{\pi }{p}\right)\right)+\frac{\pi \ln \left(\tan \left(\frac{\pi }{p}\right)\right)}{p}$ $\endgroup$ – Mariusz Iwaniuk Jul 22 at 14:25

Here is my attempt, however, it does not lead to a nice close form solution. Let $I_{p}$ be defined as follows:

\begin{equation} I_{p}=\int\limits_{0}^{\frac{\pi}{p}} \ln(\tan(x))\,dx \end{equation}

for some real valued $p$ such that $p\geq4$. With the substitution $x=\arctan(t)$, you can transform the integral to the following:

\begin{equation} I_{p}=\int\limits_{0}^{\tan(\pi/p)} \frac{\ln(t)}{1+t^{2}}\,dt \end{equation}

Now, by letting $t=e^{-z}$, you will get the following:

\begin{equation} I_{p}=\int\limits_{\ln(\cot(\pi/p))}^{+\infty} \frac{(-z)e^{-z}}{1+e^{-2z}}\,dz \end{equation}

\begin{equation} I_{p}=\int\limits_{\ln(\cot(\pi/p))}^{+\infty} \frac{(-z)e^{-z}}{1-(-e^{-2z})}\,dz \end{equation}

For simplicity, let $k=\ln(\cot(\pi/p))$. For any $p\geq4$, in the interval $[k,\infty)$, it holds that $0\leq|-e^{-2z}|\leq 1$, so it is justified to use the geometric series. Thus:

\begin{equation} I_{p}=-\int\limits_{k}^{+\infty} ze^{-z}\sum_{n=0}^{+\infty}(-e^{-2z})^{n}\,dz \end{equation}

\begin{equation} I_{p}=-\sum_{n=0}^{+\infty}(-1)^{n}\int\limits_{k}^{+\infty} ze^{-z}e^{-2nz}\,dz \end{equation}

\begin{equation} I_{p}=-\sum_{n=0}^{+\infty}(-1)^{n}\int\limits_{k}^{+\infty} ze^{-z(1+2n)}\,dz \end{equation}

By letting $s=z(1+2n)$, we obtain that:

\begin{equation} I_{p}=-\sum_{n=0}^{+\infty}\frac{(-1)^{n}}{(1+2n)^{2}}\int\limits_{k(1+2n)}^{+\infty} se^{-s}\,ds \end{equation}

Using the upper incomplete gamma function, we get the following:

\begin{equation} I_{p}=-\sum_{n=0}^{+\infty}\frac{(-1)^{n}}{(1+2n)^{2}}\Gamma(2,k+2nk) \end{equation}

\begin{equation} \boxed{\int\limits_{0}^{\frac{\pi}{p}} \ln(\tan(x))\,dx=-\sum_{n=0}^{+\infty}\frac{(-1)^{n}}{(1+2n)^{2}}\Gamma\left(2,\ln\left(\cot\left(\frac{\pi}{p}\right)\right)+2n\ln\left(\cot\left(\frac{\pi}{p}\right)\right)\right)} \end{equation}

| cite | improve this answer | |

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.