I need help to find the numerical values to the precision at least $50$ digits (the closed forms if possible) for the following integrals
\begin{equation} {\large\mathscr{F}}\left(\alpha,\beta,\mu\right)=\int_0^{\Large\frac{\pi}{2}}\bigg[\tan x\arctan\big(\beta\tan (\mu\tan x)\big)-\tan x\arctan\big(\alpha\tan (\mu\tan x)\big)\bigg]\ dx\\ \end{equation}
and
\begin{equation} {\large\mathscr{I}}=\int_0^{\Large\frac{\pi}{2}}\cot\left(\frac{\cot x}{2}\right)\cot x\ dx\\ \end{equation}
Somehow, my Mathematica $9.0$ failed to find the numerical values. It showed up warning messages like these:
NIntegrate::slwcon: Numerical integration converging too slowly; suspect one of the following: singularity, value of the integration is 0, highly oscillatory integrand, or WorkingPrecision too small. >>
or
NIntegrate::ncvb: NIntegrate failed to converge to prescribed accuracy after 9 recursive bisections in x near {x} =
I use this code to obtain the numerical value of an integral (perhaps you have a better code that you might want to share with me)
N[Integrate[(integrand), {x,a,b}], (digits precision)]
I am interested in knowing the numerical values of ${\large\mathscr{F}}\left(\alpha,\beta,\mu\right)$ for the specific values of the following variables: \begin{array}{|c|c|c|c|} \hline \text{No.} & \alpha & \beta & \mu \\[7pt] \hline 1 & 2 & 3 & 1 \\[7pt] \hline 2 & 3 & 5 & 2 \\[7pt] \hline 3 & \frac{3}{2} & 2 & \frac{1}{2}\\[7pt] \hline 4 & \frac{4}{3} & \frac{5}{3} & \frac{1}{3}\\[7pt] \hline 5 & \frac{5}{4} & \frac{3}{2} & \frac{1}{4}\\[7pt] \hline \end{array} Any help would be greatly appreciated. Thank you.
Edit :
As requested by Mr. Amzoti, I used this code (I use example values no. $1$ in the table)
N[Integrate[Tan[x] ArcTan[3 Tan[Tan[x]]] - Tan[x] ArcTan[2 Tan[Tan[x]]],{x,0,Pi/2}], 50]
For the second integral $$ \int_0^{\pi/2} \tan x\tan(2/\tan x)\,dx, $$ change the variables $$ \tan x = u, \qquad u = \frac1v $$ to get $$ \int_0^\infty \frac{dv}{v(1+v^2)}\tan 2v. $$ Then split the integration over the individual periods of $\tan$, getting the sum $$ \int_0^{\pi/4}ds\,\cot 2s \sum_{k\geq0}\left(\frac{1}{v_-(1+v_-^2)} - \frac{1}{v_+(1+v_+^2)} \right), \qquad v_\pm = \frac\pi2(k+\tfrac12)\pm s. $$ The sum can be done with a CAS in terms of polygamma functions, then the integral can be done numerically: $$ \begin{array}{rl} 1.5142909817&1206622699&9814154016&3412884691&1754296527&\\0554534308&1840600184&0711274192&8238561012&4123287312&\\7970541870&9009506626&9851358610&7548703861&4382369398&\ldots \end{array} $$
For the modified second integral using the same approach I get $$ 1.8283334379\ 3552390329\ 6350583941\ 3012496783\ 6378207215\ 1909448383\ldots $$
For the first integral, change the variable to $$ u = \tan x, \qquad u = \frac1\mu \left( \tfrac\pi2 + \pi s\right), \qquad s\geq-\tfrac12, $$ and use $$ \frac{u}{1+u^2} = \Re \frac{1}{i+u}, $$ so that the each individual part for $(\beta,\mu)$ can symbolically be written as the real part of $$ -\int_{-1/2}^\infty \frac\pi\mu\frac{\arctan(\beta\cot\pi s)}{i+u}\,ds. $$ Since $\cot\pi s$ has period $1$, split off the integral over $[-\tfrac12,0]$, and write the rest as a sum, regularized by the factor $z^k$, $|z|<1$, where the free parameter $z$ will be later taken as limiting $z\to 1$: $$ -\int_{-1/2}^0(\cdots) - \sum_{k\geq0}\int_0^1ds\,\arctan(\beta\cot\pi s)\frac{z^k \pi/\mu}{i+u+\pi k/\mu}. $$ Doing the sum over $k$, this is equal to $$ -\int_{-1/2}^0(\cdots) - \int_0^1 ds\,\arctan(\beta\cot\pi s)\Phi(z, 1, i\mu/\pi+\tfrac12+s), $$ where $\Phi$ is the Hurwitz Lerch phi function and using the limiting form $$ \Phi(z, 1, a) = -\gamma-\log(1-z)-\psi(a) + o(1), $$ ($\gamma$ is Euler's gamma, $\psi$ is digamma), the non-singular part of the integral is $$ -\int_{-1/2}^0(\cdots) - \int_0^1 ds\,\arctan(\beta\cot\pi s)(-\gamma-\Re\psi(i\mu/\pi+\tfrac12+s)) \qquad\qquad(*) $$ When two integrals $(\beta,\mu)$ and $(\alpha,\mu)$ are subtracted, the singular parts (in $z$, diverging as $\log(1-z)$ as $z\to1$) are identical: $$ \int_0^1\arctan(\beta\cot\pi s)\,ds = 0,$$ so only the non-singular parts need to be kept, which justifies this procedure. Even the $-\gamma$ term can be dropped.
The final integrals is easy to evaluate numerically as differences in $(*)$ for different values $(\beta,\mu)-(\alpha,\mu)$: $$ \begin{array}{rl} 0.0375498037&2982212925&4568723544&0398706302&0150535939&0201026802\ldots\\ 0.0048468107&7351794278&3854803548&8002752919&3912230233&5799187527\ldots\\ 0.0854502616&2875157796&1693832958&8743830124&3474549771&5684466950\ldots\\ 0.0961295058&0450574013&5986068340&3132551900&1795523024&9411717322\ldots\\ 0.0935379078&0619519591&4784481067&9022768695&3615022581&3494639786\ldots \end{array}$$