This post, after a complicated analysis, evaluates the integral $$I=\int_0^1\frac{\ln^2(x)\,\ln^3(1+x)}xdx$$

simply as

$$I =-\frac{\pi^6}{252}-18\zeta(\bar{5},1)+3\zeta^2(3)\tag1$$


$$\zeta(\bar{5},1)=\frac{1}{24}\int^1_0\frac{\ln^4{x}\ln(1+x)}{1+x}{\rm d}x$$

More succinctly,

$$I = -12\,S_{3,3}(-1)\tag2$$

with Nielsen generalized polylogarithm $S_{n,p}(z)$.

Question: How do we show that $\zeta(\bar{5},1)$ is also a Nielsen generalized polylogarithm in disguise? More generally, for $-1\leq z\leq1$, how to show

$$\begin{aligned}S_{n,p}(z) &= C_1\int_0^1\frac{(\ln x)^{n-1}\big(\ln(1-z\,x)\big)^p}{x}dx\\ &\overset{?}= C_2\int_0^1\frac{(\ln x)^{n}\;\big(\ln(1-z\,x)\big)^{p-1}}{1-z\,x}dx\end{aligned}\tag3$$

where, $$C_1 = \frac{(-1)^{n+p-1}}{(n-1)!\,p!},\qquad C_2 = \frac{(-1)^{n+p-1}}{n!\,(p-1)!}\color{red}z$$

If true, this implies,

$$\zeta(\bar{5},1) \overset{\color{red}?}= S_{4,2}(-1)\tag4$$

Edit: It turns out the notation $\zeta(\bar{5},1)$ is a multiple zeta function so,

$$\zeta(\bar{a},1)=\sum_{n=1}^{\infty}\frac{H_n}{(n+1)^a}\,(-1)^{n+1} = S_{a-1,2}(-1)$$

with harmonic numbers $H_n$, hence $(4)$ indeed is true and is just the case $a=5$. However, $(3)$ still needs to be proved in general.

  • $\begingroup$ The integral then relates two Nielsen polylogs as, $$I = -12\,S_{3,3}(-1) = -\frac{\pi^6}{252}-18\,S_{4,2}(-1)+3\zeta^2(3)$$ $\endgroup$ – Tito Piezas III Jun 3 '19 at 8:32

Without knowing much about the context for this problem, the relationship between the two integrals seems to be pretty direct from integration by parts. For $n, p \geq 1$, writing $u(x) = (\ln x)^n$ and $v(x) = (\ln(1 - zx))^p$, we have $$\frac{du}{dx} = \frac{n(\ln x)^{n-1}}{x} \qquad \text{and} \qquad \frac{dv}{dx} = -\frac{pz (\ln (1 - zx))^{p-1}}{1 - zx}$$ hence using integration by parts: \begin{align*} n\int_0^1 \frac{(\ln x)^{n-1} (\ln(1 - zx))^p}{x} \,dx &= \int_0^1 \frac{du}{dx} v \,dx \\ &= (uv)|_0^1 - \int_0^1 u \frac{dv}{dx} \,dx \\ &= pz\int_0^1 \frac{(\ln x)^n (\ln (1 - zx))^{p-1}}{1 - zx} \,dx \end{align*} where the last equation holds since $u(x)v(x) \to 0$ as $x \to 0$ or $x \to 1$.

This is $(3)$, up to the factor $(-1)^{n+p-1}/n! p!$.

  • $\begingroup$ Thanks. I suspected as much. I'm just wondering why some authors use the 2nd form of the integral, when the 1st form is better known and even has a name (Nielsen polylogs). I guess they have their reasons. $\endgroup$ – Tito Piezas III Jun 4 '19 at 2:06

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.