38

I am trying to calculate $$ I:=\int_0^1 dx \frac{\ln x \ln^2(1-x)\ln(1+x)}{x}$$ Note, the closed form is beautiful (yes beautiful) and is given by

$$ I=−\frac{3}{8}\zeta_2\zeta_3 -\frac{2}{3}\zeta_2\ln^3 2 +\frac{7}{4}\zeta_3\ln^2 2-\frac{7}{2}\zeta_5+4\ln 2 \operatorname{Li}_4\left(\frac{1}{2}\right)+\frac{2}{15}\ln^5 2+4\operatorname{Li}_5\left(\frac{1}{2}\right) $$ where $$ \zeta_s=\sum_{n=1}^\infty \frac{1}{n^{s}},\qquad \operatorname{Li}_s(z)=\sum_{n=1}^\infty \frac{z^n}{n^s},\qquad\text{for}\ |z|<1. $$ I succeeded in writing the integral as $$ I=-\sum_{i=0}^\infty \int_0^1 x^i\ln x\ln(1+x)\ln(1-x)\ dx, $$ but I am confused as to where to go from here. Possibly I was thinking of trying to use Mellin transforms or residues.

A reference to aid us is here. (Since somebody has asked for reference)

We can also write I as $$ I=\sum_{i=0}^\infty \sum_{j=1}^\infty \frac{1}{j}\sum_{k=1}^\infty \frac{1}{k} \int_0^1 x^{i+j+k} \ln x\ dx $$ using $$ \int_0^1 x^n \ln x\ dx= -\frac{1}{(n+1)^2}, $$ we can simplify this, but I am not sure then how to compute the triple sum. Thank you again.

Machinato
  • 3,207
Jeff Faraci
  • 10,393
  • 10
    Out of curiosity, what makes this closed form particularly beautiful? – Alex Becker Apr 20 '14 at 16:56
  • 3
    No doubt: "beautiful" is in the eyes of the beholder...If only the Riemann zeta function appeared I'd agree, but those$;\color{red}L$ogarithmic $;\color{red}i$ntegrals trash, imo, the beauty. – DonAntonio Apr 20 '14 at 16:58
  • @AlexBecker I would rather not discuss the form of beauty here, as we are not in a greek philosophy reading the Republic. However, this is a very nice integral from a mathematics journal that I found very beautiful. I apologize if you don't like it as much as I do. – Jeff Faraci Apr 20 '14 at 17:02
  • This closed form looks very much as if it were determined via PSLQ? Is this the case? – Eric Towers Apr 20 '14 at 17:03
  • @DonAntonio Yes beauty is in the eye of the beholder! – Jeff Faraci Apr 20 '14 at 17:03
  • @EricTowers No this is from a mathematics journal. The closed form is from there. – Jeff Faraci Apr 20 '14 at 17:03
  • Could we trouble you for a reference in the question? – Eric Towers Apr 20 '14 at 17:04
  • @EricTowers do you want the reference? Ok I just posted it for you, no i'm not being intentionally vague. What is with all the rude comments, I'm just trying to post interesting integrals that challenge the great mathematicians on this website. I am really baffled by all the rude comments. – Jeff Faraci Apr 20 '14 at 17:04
  • Don't take it hard, @Integrals: sometimes we get a little astray because of not so-mathematical contents in some question/answer. THe beautiful part is unimportant and all the comments were kidding/in a joyful fashion. What is a little more important is the lack of proper references and self work, but now you've covered that, too. – DonAntonio Apr 20 '14 at 17:10
  • @DonAntonio In every single post I ever post, I provide self work and ALWAYS references if people ask. Please see all of my past questions. I also think it is quite obvious that I purely post problems to interest the great people on this site, and they have nothing to do with "homework". – Jeff Faraci Apr 20 '14 at 17:11
  • 6
    Since the polylogs are being evaluated at $\frac{1}{2}$, I don't feel that the beauty of the answer is ruined. – Random Variable Apr 20 '14 at 17:47
  • @RandomVariable Thanks for the positive support on posting this question!! As always. And I do agree with you. – Jeff Faraci Apr 20 '14 at 17:50
  • 1
    Sos440 gave an evaluation (which I helped simplify a bit) for the case where $\ln(1+x)$ is squared instead of $\ln(1-x)$. http://math.stackexchange.com/questions/560950/how-to-evaluate-int-01-log-x-log1-x-log21x-over-x-dx – Random Variable Apr 20 '14 at 18:08
  • @RandomVariable VERY NICE! It is interesting, and I didn't think of this as I haven't seen it in the paper. Thanks a lot. The result seems a bit more lengthy for the case of $\ln^2(1-x)$ however both look very nice closed form – Jeff Faraci Apr 20 '14 at 21:49

3 Answers3

26

We will use similar approach as sos440's answer in I&S. Using the simple algebraic identity $$ ab^2=\frac{(a+b)^3+(a-b)^3-2a^3}{6}, $$ it follows that \begin{align} \int_0^1 \frac{\ln x\ln(1+x)\ln^2(1-x)}{x}\ dx &=\frac16I_1+\frac16I_2-\frac13I_3\ ,\tag1 \end{align} where \begin{align} I_1&=\int_0^1\frac{\ln x\ln^3(1-x^2)}{x}\ dx\\[12pt] I_2&=\int_0^1\frac{\ln x}{x}\ln^3\left(\frac{1+x}{1-x}\right)\ dx\\[12pt] I_3&=\int_0^1\frac{\ln x\ln^3(1+x)}{x}\ dx \end{align}


Evaluation of $I_1$

Setting $t=x^2$ followed by $t\mapsto1-t$, we have \begin{align} I_1&=\frac14\int_0^1\frac{\ln t\ln^3(1-t)}{t}\ dt\\ &=\frac14\int_0^1\frac{\ln (1-t)\ln^3t}{1-t}\ dt\\ \end{align} To evaluate the integral above, we can use multiple derivative of beta function

$$ I_1=\frac14\lim_{x\to1}\lim_{y\to0^+}\frac{\partial^4\text{B}(x,y)}{\partial x^3\partial y}=3\zeta(5)-\frac32\zeta(2)\zeta(3).\tag2 $$

Alternatively, we can use generating function for the harmonic numbers for $|z|<1$ $$ \sum_{n=1}^\infty H_n z^n=-\frac{\ln(1-z)}{1-z}, $$ identity of the harmonic numbers $$ H_{n+1}-H_n=\frac1{n+1}, $$ and $$ \int_0^1 x^\alpha \ln^n x\ dx=\frac{(-1)^n n!}{(\alpha+1)^{n+1}}, \qquad\text{for }\ n=0,1,2,\ldots\tag3 $$ We may refer to the following answer to see the complete approach for evaluating $I_1$.


Evaluation of $I_2$

$I_2$ has been evaluated by sos440 here and it is equal to

$$ I_2=\frac{21}{4}\zeta(2)\zeta(3)-\frac{93}{8}\zeta(5).\tag4 $$

Alternatively, we can use the following technique. Setting $t=\dfrac{1-x}{1+x}\ \color{red}{\Rightarrow}\ x=\dfrac{1-t}{1+t}$ and $dx=-\dfrac{2}{(1+t)^2}\ dt$, then \begin{align} I_2&=-\int_0^1\frac{\ln x}{x}\ln^3\left(\frac{1-x}{1+x}\right)\ dx\\ &=2\int_0^1\frac{\ln^3 t\ln(1+t)}{(1-t)(1+t)}\ dt-2\int_0^1\frac{\ln^3 t\ln(1-t)}{(1-t)(1+t)}\ dt.\tag5 \end{align} Using the fact that $$ \frac{2}{(1-t)(1+t)}=\frac1{1-t}+\frac1{1+t} $$ and $(5)$ can be evaluated by performing some tedious calculations involving series expansion (double summation or generating function for the harmonic numbers) of the form $\dfrac{\ln(1\pm t)}{1\pm t}$ and equation $(3)$.

Another alternative way to evaluate $I_2$ without using complex analysis and dividing integral into four separated integrals is substituting $t=\dfrac{1-x}{1+x}$ and $I_2$ turns out to be $$ I_2=-2\int_0^1\frac{\ln^3t}{1-t^2}\ln\left(\frac{1-t}{1+t}\right)\ dt,\tag6 $$ where $(6)$ has been evaluated by Omran Kouba (see evaluation of $K$).


Evaluation of $I_3$

$I_3$ has been evaluated here and it is equal to

\begin{align} I_3=&\ \frac{\pi^2}2\zeta(3)+\frac{99}{16}\zeta(5)-\frac25\ln^52+\frac{\pi^2}3\ln^32-\frac{21}4\zeta(3)\ln^22\\&-12\operatorname{Li}_4\left(\frac12\right)\ln2-12\operatorname{Li}_5\left(\frac12\right).\tag7 \end{align}


Thus, putting altogether we obtain

$$ I=\color{blue}{\small{\frac{2}{15}\ln^5 2-\frac{2}{3}\zeta(2)\ln^3 2 +\frac{7}{4}\zeta(3)\ln^2 2−\frac{3}{8}\zeta(2)\zeta(3) -\frac{7}{2}\zeta(5)+4\operatorname{Li}_4\left(\frac12\right)\ln2+4\operatorname{Li}_5\left(\frac{1}{2}\right)}}. $$


$$ \large\color{blue}{\text{# }\mathbb{Q.E.D.}\text{ #}} $$

Tunk-Fey
  • 20,650
2

using the following identity proved by Cornel and can be found in his book, (Almost) Impossible Integrals, Sums and Series. $\quad\displaystyle\ln(1-x)\ln(1+x)=-\sum_{n=1}^\infty\left(\frac{H_{2n}-H_n}{n}+\frac1{2n^2}\right)x^{2n}$.

multiply both sides by $\frac{\ln(1-x)}{x}$ then integrate from $x=0$ to $1$, we get \begin{align} I&=-\sum_{n=1}^\infty\left(\frac{H_{2n}-H_n}{n}+\frac1{2n^2}\right)\int_0^1x^{2n-1}\ln(1-x)\ dx\\ &=-\sum_{n=1}^\infty\left(\frac{H_{2n}-H_n}{n}+\frac1{2n^2}\right)\left(-\frac{H_{2n}}{2n}\right)\\ &=2\sum_{n=1}^\infty\frac{H_{2n}^2}{(2n)^2}-\frac12\sum_{n=1}^\infty\frac{H_nH_{2n}}{n^2}+\sum_{n=1}^\infty\frac{H_{2n}}{(2n)^3}\\ &=\sum_{n=1}^\infty\frac{H_{n}^2}{n^2}+\sum_{n=1}^\infty\frac{(-1)^nH_{n}^2}{n^2}-\frac12\color{blue}{\sum_{n=1}^\infty\frac{H_{n}H_{2n}}{n^2}}+\sum_{n=1}^\infty\frac{H_{n}}{n^3}+\sum_{n=1}^\infty\frac{(-1)^nH_{n}}{n^3} \end{align} I managed here to prove $$\color{blue}{\sum_{n=1}^\infty\frac{H_nH_{2n}}{n^2}}=4\sum_{n=1}^\infty\frac{H_n^2}{n^2}+2\sum_{n=1}^\infty\frac{(-1)^nH_n^2}{n^2}+2\sum_{n=1}^\infty\frac{(-1)^nH_n}{n^3}-4\sum_{n=1}^\infty\frac{H_n^{(3)}}{n2^n}-6\zeta(4)$$ which follows that $$\boxed{I=\sum_{n=1}^\infty\frac{H_n}{n^3}-\sum_{n=1}^\infty\frac{H_n^2}{n^2}+\sum_{n=1}^\infty\frac{H_n^{(3)}}{n2^n}+3\zeta(4)}$$ Plugging the following results: $$\displaystyle\sum_{n=1}^{\infty}\frac{H_n^2}{n^2}=6\zeta(4)-\frac74\zeta(4)=\frac{17}4\zeta(4)$$

$$\sum_{n=1}^\infty\frac{H_n^{(3)}}{n2^n}=\operatorname{Li_4}\left(\frac12\right)-\frac{5}{16}\zeta(4)+\frac78\ln2\zeta(3)-\frac14\ln^22\zeta(2)+\frac{1}{24}\ln^42$$

The first sum can be found here, and the second sum can be found using the easy-to-prove generating function $\sum_{n=1}^\infty\frac{x^nH_n^{(3)}}{n}=\operatorname{Li_4}(x)-\ln(1-x)\operatorname{Li_3}(x)-\frac12\operatorname{Li_2}^2(x)$,substituting these two sums along with the well known value of $\sum_{n=1}^\infty\frac{H_n}{n^3}=3\zeta(5)-\zeta(2)\zeta(3)$, we get the desired closed form of $I$.

NOTE . Sorry guys I just noticed that my solution is for $\int_0^1\frac{\ln^2(1-x)\ln(1+x)}{x}\ dx$ without $\ln x$ in the numerator as in the original problem. I am keeping the solution as it was voted as useful.

Ali Olaikhan
  • 27,891
1

Using the same start of Tunk-Fey's solution:

$$\begin{align} I=\int_0^1 \frac{\ln x\ln(1+x)\ln^2(1-x)}{x}\ dx &=\frac16I_1-\frac16I_2-\frac13I_3 \end{align}$$

where

\begin{align} I_1&=\int_0^1\frac{\ln x\ln^3(1-x^2)}{x}\ dx\\[12pt] I_2&=\int_0^1\frac{\ln x}{x}\ln^3\left(\frac{1-x}{1+x}\right)\ dx\\[12pt] I_3&=\int_0^1\frac{\ln x\ln^3(1+x)}{x}\ dx \end{align} What makes my solution different is that i am going to evaluate these integrals in different ways.


\begin{align} I_1&=\int_0^1\frac{\ln x\ln^3(1-x^2)}{x}\ dx=\frac14\int_0^1\frac{\ln x\ln^3(1-x)}{x}\ dx\\ &=\int_0^1\frac{\ln(1-x)\ln^3x}{1-x}\ dx=-\sum_{n=1}^\infty H_n\int_0^1x^n\ln^3x\ dx\\ &=6\sum_{n=1}^\infty\frac{H_n}{(n+1)^4}=6\sum_{n=1}^\infty\frac{H_n}{n^4}-6\zeta(5)=\boxed{12\zeta(5)-6\zeta(2)\zeta(3)=I_1} \end{align}

where we substituted $\sum_{n=1}^\infty\frac{H_n}{n^4}=3\zeta(5)-\zeta(2)\zeta(3)$


$$I_2=\small{\int_0^1\frac{\ln x}{x}\ln^3\left(\frac{1-x}{1+x}\right)\ dx\overset{IBP}{=}3\int_0^1\frac{\ln^2x}{1-x^2}\ln^2\left(\frac{1-x}{1+x}\right)\ dx\overset{\frac{1-x}{1+x}=y}{=}\frac32\int_0^1\frac{\ln^2x}{x}\ln^2\left(\frac{1-x}{1+x}\right)\ dx}$$

and by using $\ln^2\left(\frac{1-x}{1+x}\right)=-2\sum_{n=1}^\infty\frac{H_n-2H_{2n}}{n}x^{2n}$ (proved here) we get \begin{align} I_2&=-3\sum_{n=1}^\infty\frac{H_n-2H_{2n}}{n}\int_0^1x^{2n-1}\ln^2x\ dx=24\sum_{n=1}^\infty\frac{H_{2n}}{(2n)^4}-3\sum_{n=1}^\infty\frac{H_n}{n^4}\\ &=12\sum_{n=1}^\infty\frac{(-1)^nH_n}{n^4}+\frac{45}4\sum_{n=1}^\infty\frac{H_n}{n^4}=\boxed{\frac{93}{8}\zeta(5)-\frac{21}{4}\zeta(2)\zeta(3)=I_2} \end{align} where we substituted $\sum_{n=1}^\infty\frac{(-1)^nH_n}{n^4}$$=\frac12\zeta(2)\zeta(3)-\frac{59}{32}\zeta(5)\ $


$I_3$ was solved here in different ways but this solution is my favorite one which gives

$$\boxed{\small{I_3=-12\operatorname{Li}_5\left(\frac12\right)-12\ln2\operatorname{Li}_4\left(\frac12\right)+\frac{99}{16}\zeta(5)+3\zeta(2)\zeta(3)-\frac{21}4\ln^22\zeta(3)+2\ln^32\zeta(2)-\frac25\ln^52}\ \ }$$


Combine $I_1$, $I_2$ and $I_3$ to get

$$I=\small{4\operatorname{Li}_5\left(\frac{1}{2}\right)+4\ln2\operatorname{Li}_4\left(\frac12\right)-\frac{7}{2}\zeta(5)−\frac{3}{8}\zeta(2)\zeta(3)+\frac{7}{4}\ln^22\zeta(3)-\frac{2}{3}\ln^32\zeta(2)+\frac{2}{15}\ln^5 2}$$

Ali Olaikhan
  • 27,891