1
$\begingroup$

I am now trying a direct approach to solving my question about $$\int_0^\infty\frac{\arctan a_1x\arctan a_2x\dots\arctan a_nx}{1+x^2}\,dx$$ where the $a_i$ are all positive. Note that the $\arctan$s can be broken up into logarithms and that the denominator is $(1+ix)(1-ix)$. Now this answer claims at the end

Any integral of the form $\int\frac{\ln(ax+b)\ln(cx+d)}{px+q}\,dx$ can be systematically reduced to trilogarithms, dilogarithms, and elementary functions.

Extrapolating, I am led to believe that any integral of the form $$\int\frac{\log(a_1x+b_1)\cdots\log(a_nx+b_n)}{px+q}\,dx$$ can be systematically reduced to elementary functions and polylogarithms of order up to $n+1$. This would (theoretically) solve my problem.

Is my statement true and if so how would I write an algorithm to perform the polylog reduction?

$\endgroup$
1
  • $\begingroup$ This is the question that led me into a deep dive into polylogarithms 10 years ago, and unfortunately the answer is in the negative, at least as far as anyone knows. Even the question for the product of three logs in the numerator is unsettled $\endgroup$
    – David H
    Commented Apr 19 at 5:11

0

You must log in to answer this question.

Browse other questions tagged .