6

(Motivation) In an attempt to answer this question, I got stuck on evaluating a certain integral. I have made up the following conjecture:

$$\int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx = \frac{\pi^{2}}{4}+\frac{3}{2}\ln\left(2\right)\ln\left(\frac{\sqrt{3}+1}{\sqrt{3}-1}\right).$$

(Attempt) Let $H_n$ denote the n-th harmonic number $\displaystyle H_{n}=\sum_{k=1}^{n}\frac{1}{k}$. I will warn this process gets ugly, but it is the best I have so far. Expanding the integrand as a series, we get:

$$ \eqalign{ \int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx &= \sum_{n=0}^{\infty}\int_{\sqrt{3}}^{\infty}\frac{4\ln\left(x\right)-H_{n}}{x^{4n+2}}dx+\sum_{n=0}^{\infty}\int_{\sqrt{3}}^{\infty}\frac{4\ln\left(x\right)-H_{n}}{x^{4n+4}}dx, \cr } $$

which simplifies down to

$$\sum_{n=0}^{\infty}\left(\frac{H_{n}\sqrt{3}^{-4n-1}}{-4n-1}-\frac{4\ln\left(\sqrt{3}\right)\sqrt{3}^{-4n-1}}{-4n-1}+\frac{4\sqrt{3}^{-4n-1}}{\left(-4n-1\right)^{2}}\right)+\sum_{n=0}^{\infty}\left(\frac{H_{n}\sqrt{3}^{-4n-3}}{-4n-3}-\frac{4\ln\left(\sqrt{3}\right)\sqrt{3}^{-4n-3}}{-4n-3}+\frac{4\sqrt{3}^{-4n-3}}{\left(-4n-3\right)^{2}}\right).$$

Since both series converge, we can split up the first series like

$$-\sum_{n=0}^{\infty}\frac{H_{n}\sqrt{3}^{-4n-1}}{4n+1}+4\ln\left(\sqrt{3}\right)\sum_{n=0}^{\infty}\frac{\sqrt{3}^{-4n-1}}{4n+1}+4\sum_{n=0}^{\infty}\frac{\sqrt{3}^{-4n-1}}{\left(4n+1\right)^{2}}$$

and the second series like

$$-\sum_{n=0}^{\infty}\frac{H_{n}\sqrt{3}^{-4n-3}}{4n+3}+4\ln\left(\sqrt{3}\right)\sum_{n=0}^{\infty}\frac{\sqrt{3}^{-4n-3}}{4n+3}+4\sum_{n=0}^{\infty}\frac{\sqrt{3}^{-4n-3}}{\left(4n+3\right)^{2}}.$$

Next, I found that

$$\sum_{n=0}^{\infty}\frac{\sqrt{3}^{-4n-1}}{4n+1}=\frac{\pi}{12}+\frac{1}{2}\operatorname{arctanh}\left(\frac{1}{\sqrt{3}}\right)$$

and

$$\sum_{n=0}^{\infty}\frac{\sqrt{3}^{-4n-3}}{4n+3}=\frac{1}{2}\operatorname{arctanh}\left(\frac{1}{\sqrt{3}}\right)-\frac{\pi}{12}.$$

However, after trying for a while, I am out of ideas for evaluating the other sums.

I realize I skipped a lot of steps, but that is because I don't want this question to be too long. So for your convenience, I put all of these into Desmos, so I believe my process is correct so far based on numerical approximations.

(Question) Does anyone have an idea of how to evaluate the integral in question, or how to evaluate the sums I am stuck on? Any hints and ideas are appreciated.

(Miscellaneous) Here are some other ideas I have:

$$ \eqalign{ \int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx &= \int_{\operatorname{arcsec}\left(\sqrt{3}\right)}^{\frac{\pi}{2}}\frac{\ln\left(\left(\sec x\right)^{4}-1\right)}{\sec^{2}x-1}\sec\left(x\right)\tan\left(x\right)dx \cr \int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx &= \int_{0}^{\infty}\frac{\ln\left(\left(x+\sqrt{3}\right)^{4}-1\right)}{\left(x+\sqrt{3}\right)^{2}-1}dx. } $$

Maybe I could construct a keyhole contour for the last integral?

Accelerator
  • 4,923
  • 2
    In regards to your last remark, keyhole would be rather hard. Your integrand is quite exotic, and frankly I’m not sure what I would do with it in a complex analysis setting. There are four branch points, so a keyhole would need to be… a quadruple keyhole? Or you probably can do something with the branch cuts to make them cancel in some way? Plus, the poles lie along the branch cuts which make me very unsure of what to do with them lol – Captain Chicky Nov 14 '22 at 06:17
  • @CaptainChicky I found there is a branch point at $x=1-\sqrt{3}$, so I thought about constructing a rectangular contour on the first quadrant, or a contour shaped like one-fourth of a pizza on the first quadrant. But yes, those would be very tricky. – Accelerator Nov 14 '22 at 06:27
  • 1
    WolframAlpha seems happy to do the indefinitie integral in terms of the dilogarithm. There are also some ideas which may be useful here: https://math.stackexchange.com/questions/177160/integral-int-infty-infty-frac-lnx21x21dx – Blitzer Nov 14 '22 at 06:31
  • 2
    Well I bruteforced this using matematica and got this
       4 (2 Log[-1 + Sqrt[3]]^2 + 4 Log[1 + Sqrt[3]]^2 + 
          Log[2] Log[18817/2 + 5432 Sqrt[3]] + 
          4 (PolyLog[2, 1/2 (1 - Sqrt[3])] + 
             PolyLog[2, (-(1/2) - I/2) (-1 + Sqrt[3])] + 
             PolyLog[2, (-(1/2) + I/2) (-1 + Sqrt[3])] + 
             PolyLog[2, -1 + Sqrt[3]] + 
             PolyLog[2, (1 - I)/(1 + Sqrt[3])] + 
             PolyLog[2, (1 + I)/(1 + Sqrt[3])])))```
    
    – Max0815 Nov 14 '22 at 06:37
  • 1
    I have tested this value numerically to your conjectured closed form and it holds to at least 1000 decimals. I suspect that this integral in and of itself is another complicated polylog identity bash like the original integral that prompted you to solve this one in the first place. BTW, what motivated your conjectured closed form? – Max0815 Nov 14 '22 at 06:39
  • 1
    @Max0815 While working on the integral in the link at the beginning which was the conjecture $\int_{0}^{1}\frac{\ln\left(\frac{x+1}{2x^{2}}\right)}{\sqrt{x^{2}+2x}}dx = \frac{\pi^{2}}{2}$, I was able to prove that $\int_{0}^{1}\frac{\ln\left(\frac{x+1}{2x^{2}}\right)}{\sqrt{x^{2}+2x}}dx = 2\int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx-3\ln\left(2\right)\ln\left(\frac{\sqrt{3}+1}{\sqrt{3}-1}\right)$ after a bunch of tedious work, so then I set it equal to $\frac{\pi^{2}}{2}$ and solved for $\int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx$. – Accelerator Nov 14 '22 at 06:44
  • Your conjecture is equivalent to the integral $\int_{0}^{1}\frac{\ln\left(\frac{x+1}{2x^{2}}\right)}{\sqrt{x^{2}+2x}}dx = \frac{\pi^{2}}{2}$, which has already been proved. – pisco Nov 14 '22 at 14:35
  • Do you have the link for the proof? @pisco – Accelerator Nov 14 '22 at 19:00
  • @Accelerator: the link is in the question. The link between these two integrals is a change of variable or two. – FDP Nov 14 '22 at 20:29
  • @Accelerator It has already been proved in the answers to that question. See for example, answers of FDP and mine. – pisco Nov 14 '22 at 20:51
  • There must be a very easy way. Nobody sees it. – Bob Dobbs Nov 16 '22 at 11:04
  • Late response, but I read your proof and I understood it. Thank you. I should've read it a bit more thoroughly back when I posted this query. @pisco – Accelerator Nov 18 '22 at 11:16
  • @BobDobbs I'm not sure if my answer down below is an easy way, but I would say once you know about that certain dilogarithm derivation I cited in my answer, then it's a lot easier. – Accelerator Nov 18 '22 at 22:39
  • @Accelerator Yes but you used an answer and what if this answer used another answer, am I going to click all these links all day? – Bob Dobbs Nov 19 '22 at 08:29
  • 1
    Most of my work in my answer was made up by me, and I just used a derivation to help me with the last part. Part of that derivation came from another previous derivation. Is there something wrong with that? @BobDobbs – Accelerator Nov 19 '22 at 21:40
  • @Accelerator I know. I was just joking. Sorry. – Bob Dobbs Nov 20 '22 at 07:11

4 Answers4

3

I will answer my own question by proving

$$\int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx = \frac{\pi^{2}}{4}+\frac{3}{2}\ln\left(2\right)\ln\left(2+\sqrt{3}\right).$$

Let the integral in question equal $I$. Letting $x \to \dfrac{1-x}{1+x}$, we rewrite the integral as

$$I = \int_{\sqrt{3}-2}^{-1}\frac{\ln\left(\left(\frac{1-x}{1+x}\right)^{4}-1\right)}{\left(\frac{1-x}{1+x}\right)^{2}-1}\left(\frac{-2}{\left(1+x\right)^{2}}\right)dx.$$

Doing some simplifications, we get

$$I = \frac{3\ln\left(2\right)}{2}\int_{\sqrt{3}-2}^{-1}\frac{1}{x}dx+\frac{1}{2}\int_{\sqrt{3}-2}^{-1}\frac{\ln\left(-x\right)}{x}dx+\frac{1}{2}\int_{\sqrt{3}-2}^{-1}\frac{\ln\left(1+x^{2}\right)}{x}dx-\frac{1}{2}\int_{\sqrt{3}-2}^{-1}\frac{\ln\left(\left(x+1\right)^{4}\right)}{x}dx.$$

Trivially, we can solve the first two integrals. For the last two, we can use the dilogarithm definition. Thus,

$$I = \frac{3\ln\left(2\right)}{2}\left(-\ln\left(2-\sqrt{3}\right)\right)+\frac{1}{2}\left(-\frac{1}{2}\ln^{2}\left(2-\sqrt{3}\right)\right)+\frac{1}{2}\left(\frac{1}{24}\left(12\operatorname{Li}_2 \left(4\sqrt{3}-7\right)+\pi^{2}\right)\right)-\frac{1}{2}\left(4\left(\operatorname{Li}_2 \left(2-\sqrt{3}\right)-\frac{\pi^{2}}{6}\right)\right)$$

which simplifies down to

$$I = \ln\left(\frac{1}{2-\sqrt{3}}\right)\left(\frac{1}{4}\ln\left(2-\sqrt{3}\right)+\frac{3}{2}\ln\left(2\right)\right)+\frac{17\pi^{2}}{48}+\frac{1}{4}\operatorname{Li}_2 \left(-\left(2-\sqrt{3}\right)^{2}\right)-2\operatorname{Li}_2 \left(2-\sqrt{3}\right)$$

From this answer, we can prove that

$$\frac{1}{4}\operatorname{Li}_2 \left(-\left(2-\sqrt{3}\right)^{2}\right)-2\operatorname{Li}_2 \left(2-\sqrt{3}\right) = \frac{\ln^{2}\left(2-\sqrt{3}\right)}{4}-\frac{5\pi^{2}}{48}.$$

Combining the results, we conclude that the integral $I$ is

$$\int_{\sqrt{3}}^{\infty}\frac{\ln\left(x^{4}-1\right)}{x^{2}-1}dx = \frac{\pi^{2}}{4}+\frac{3}{2}\ln\left(2\right)\ln\left(2+\sqrt{3}\right).$$

Accelerator
  • 4,923
  • 1
    @ Accelerator. Really great work! Did you try $\sqrt {2}$ instead of $\sqrt {3}$? I have found this subexpression $\frac{1}{4} \text{Li}_2\left(-\left(3-2 \sqrt{2}\right)^2\right)-2 \text{Li}_2\left(3-2 \sqrt{2}\right)$ which, however, I cannot reduce further to logs only. – Dr. Wolfgang Hintze Dec 13 '22 at 14:07
  • Thanks! I never tried that, and I have no idea how to simplify that. @Dr.WolfgangHintze – Accelerator Dec 13 '22 at 14:15
2

A possible path and a partial answer.

I would split the (indefinite) integral, first, and then take the limit. \begin{eqnarray} \mathcal I &=& \int \frac{\log(x^4-1)}{x^2-1}dx=\\ &=&\underbrace{\int\frac{\log(x-1)}{x^2-1}dx}_{\mathcal I_1} + \underbrace{\int \frac{\log(x+1)}{x^2-1}dx}_{\mathcal I_2} +\\ & &+\int \frac{\log(x+i)}{x^2-1}dx +\int \frac{\log(x-i)}{x^2-1}dx=\\ &=& \mathcal I_1 + \mathcal I_2 + \underbrace{2\mbox{Re}\left\{\int\frac{\log(x+i)}{x^2-1}dx\right\}}_{\mathcal I_3}. \end{eqnarray}

Then we have \begin{eqnarray} \mathcal I_1 &\stackrel{x-1 \mapsto t}{=}& \int \frac{\log t}{t(t+2)}dt=\\ &=&\frac12 \int \frac{\log t}t dt -\underbrace{\frac12\int \frac{\log t}{t+2}dt}_{IBP}=\\ &=&\frac14\log^2(x-1)-\frac12 \log t\log\left(\frac{t+2}2\right)-\frac12\int\frac{\log\left(1+\frac{t}2\right)}{-\frac{t}2}dt=\\ &=&\frac14 \log^2(x-1)-\frac12 \log(x-1)\log\left(\frac{x+1}2\right)-\frac12 \mbox{Li}_2 \left(-\frac{t}2\right)+C=\\ &=&\underbrace{\frac14 \log^2(x-1)-\frac12 \log(x-1)\log\left(\frac{x+1}2\right)-\frac12 \mbox{Li}_2 \left(-\frac{x-1}2\right)}_{F_1(x)}+C \end{eqnarray} Thus the definite integral gives \begin{eqnarray} \int_\sqrt 3^\infty \frac{\log(x-1)}{x^2-1}dx &=&-\frac14\log^2(\sqrt 3-1) +\frac12 \log(\sqrt 3-1)\log\left(\frac{\sqrt 3+1}2\right)+\\ & &+\frac12\mbox{Li}_2\left(-\frac{\sqrt 3-1}2\right)+\\ & &+\lim_{x\to \infty}F_1(x). \end{eqnarray} Let us first rewrite $F_1(x)$ using the reflection formula $$\mbox{Li}_2(z) = -\mbox{Li}_2\left(\frac1{z}\right)-\frac{\pi^2}6-\frac12 \log^2(-z),$$ in order to have \begin{eqnarray} F_1(x)&=& \frac14\log^2(x-1)-\frac12\log(x-1)\log\left(\frac{x+1}2\right)+\frac14\log^2\left(\frac{x-1}2\right)+\\& &+\frac{\pi^2}{12}+\frac12\mbox{Li}_2\left(-\frac2{x-1}\right)=\\ &=&\frac12\log(x-1)\log\left(\frac{x-1}{x+1}\right)+ \frac{\log^22}4+\frac{\pi^2}{12}+\frac12\mbox{Li}_2\left(-\frac2{x-1}\right). \end{eqnarray} Thus $$\lim_{x\to \infty} F_1(x) = \frac{\log^22}4 + \frac{\pi^2}{12},$$ and the first (definite) integral becomes \begin{eqnarray} \int_\sqrt 3^\infty \frac{\log(x-1)}{x^2-1}dx &=&-\frac14\log^2(\sqrt 3-1) +\frac12 \log(\sqrt 3-1)\log\left(\frac{\sqrt 3+1}2\right)+\\ & &+\frac12\mbox{Li}_2\left(-\frac{\sqrt 3-1}2\right)+\frac{\log^22}4 + \frac{\pi^2}{12}. \end{eqnarray}

An identical approach can be adopted for $\mathcal I_2$ yielding \begin{eqnarray} \int_\sqrt 3^\infty \frac{\log(x+1)}{x^2-1} dx &=&\frac14\log^2(\sqrt 3+1)-\frac12 \log 2\log(\sqrt 3-1)+\\ & &+\frac12 \mbox{Li}_2\left(-\frac{\sqrt 3-1}2\right)+\frac{\log^22}4 + \frac{\pi^2}{12}, \end{eqnarray} and therefore \begin{eqnarray} \int_\sqrt 3^\infty \frac{\log(x^2-1)}{x^2-1}dx &=&\frac14\log^2(\sqrt 3+1)-\frac14\log^2(\sqrt 3-1)- \log 2\log(\sqrt 3-1)+\\ & &+\frac{\log2}2 + \frac{\log^22}2+\mbox{Li}_2\left(-\frac{\sqrt 3-1}2\right)+\frac{\pi^2}6. \end{eqnarray}

What I have done so far for $\mathcal I_3$: \begin{eqnarray} \mathcal I_3 &=&\mbox{Re}\left\{\underbrace{\int\frac{\log(x+i)}{x-1}dx}_{IBP}-\underbrace{\int\frac{\log(x+i)}{x+1}dx}_{IBP}\right\}=\\ &=& \mbox{Re}\left\{\log\left[\left(-\frac12+\frac{i}2\right)(x-1)\right]\log(x+i)-\int\frac{\log\left[1+\left(\frac12-\frac{i}2\right)(x+i)\right]}{(x+i)}dx\right.+\\ & &\left.-\log\left[\left(\frac12+\frac{i}2\right)(x+1)\right]\log(x+i)+\int\frac{\log\left[1+\left(-\frac12-\frac{i}2\right)(x+i)\right]}{(x+i)}dx\right\}=\\ &=&\mbox{Re}\left\{\log\left[\left(-\frac12+\frac{i}2\right)(x-1)\right]\log(x+i)+\mbox{Li}_2\left[\left(\frac12-\frac{i}2\right)(x+i)\right]\right.+\\ & &\left.-\log\left[\left(\frac12+\frac{i}2\right)(x+1)\right]\log(x+i)-\mbox{Li}_2\left[\left(-\frac12-\frac{i}2\right)(x+i)\right]+C\right\} \end{eqnarray}

dfnu
  • 7,528
1

Mathematica 13.1 produces

Integrate[Log[x^4 - 1]/(x^2 - 1), {x, Sqrt[3], Infinity}]//FullSimplify

$$\frac{1}{32} \left(16 \left(\text{Li}_2\left(\frac{1}{2} \left(1-\sqrt{3}\right)\right)+\text{Li}_2\left(\sqrt{3}-1\right)+\text{Li}_2\left(\left(-\frac{1}{2}+\frac{i}{2}\right) \left(\sqrt{3}-1\right)\right)+\text{Li}_2\left(\left(-\frac{1}{2}-\frac{i}{2}\right) \left(\sqrt{3}-1\right)\right)+\text{Li}_2\left(\frac{1-i}{\sqrt{3}+1}\right)+\text{Li}_2\left(\frac{1+i}{\sqrt{3}+1}\right)+\log ^2\left(\sqrt{3}+1\right)\right)+7 \pi ^2+8 \log ^2\left(\sqrt{3}-1\right)+4 \log (2) \left(\log (8)+4 \log \left(2 \sqrt{3}+\frac{7}{2}\right)\right)\right) $$

user64494
  • 5,811
  • 2
    The hardest thing is to find out the relations between dilogs. The evaluation in terms of sum of dilogs is not hard to do for this integral. – FDP Nov 14 '22 at 20:03
1

This is not a complete solution but an extended comment.

I was particularly interested to find out what makes the lower integration border $\sqrt{3}$ so special as a prerequisite for an appreciable simplification of the final expression of the integral.

Hence I started with the same decomposition of our integral into the sum of three integrals as @dfnu, but now with a general lower integration border $a$, where $a>1$ to ensure convergence:

$$i_1(a) =\int_a^{\infty } \frac{\log (x-1)}{x^2-1} \, dx$$ $$i_2(a) =\int_a^{\infty } \frac{\log (x+1)}{x^2-1} \, dx$$

and looked particularly at the integral

$$i_3(a) = \int_{a}^{\infty}\frac{\log(x^2+1)}{x^2-1}$$

After a lengthy journey I found the following expression

$$\begin {align} i_3(a) = \frac{\pi ^2}{4}-2 \sum _{k=1}^{\infty } \frac{\left(\frac{\sqrt{a^2+1}}{\sqrt{2}}\right)^k \sin \left(\frac{\pi k}{4}\right) \sin \left(k \tan ^{-1}(a)\right)}{k^2}\\+\frac{1}{2} \log \left(\frac{a+1}{a-1}\right) \log \left(a^2+1\right)+\frac{1}{2} \pi \tan ^{-1}\left(\frac{1}{a}\right)\end{align}$$

Now to the special role of $\sqrt{3}$. The only value of $a=\sqrt{m}$ with positive integer $m$ making $\tan ^{-1}\left(\frac{1}{a}\right)$ a rational multiple of $\pi$ is $\sqrt{3}$, giving $\tan ^{-1}\left(\frac{1}{\sqrt{3}}\right)=\frac{\pi }{6}.$

The complete integral evaluates to

$$\begin{align} i(a) & = \int_{a}^{\infty}\frac {\log(x^4-1)}{x^2-1} \\ & = \frac{1}{2} \pi \tan ^{-1}\left(\frac{1}{a}\right)+\frac{\pi ^2}{3} \text{(*1*)} \\ & -\frac{1}{4} \log ^2(a-1)+\frac{1}{2} \log ^2(a+1) \\ & -\frac{1}{2} \log (2) \log (a-1)+\frac{\log ^2(2)}{4} \\ & + \frac{1}{2} \log \left(\frac{a+1}{a-1}\right) \log \left(a^2+1\right) \text{(*2*)} \\ & +\frac{1}{2} \text{Li}_2\left(\frac{1-a}{2}\right)+\frac{1}{2} \text{Li}_2\left(\frac{2}{a+1}\right) \\ & +\frac{1}{2} \left(-\text{Li}_2\left(\left(\frac{1}{2}+\frac{i}{2}\right) (a-i)\right) \\ + \text{Li}_2\left(\left(-\frac{1}{2}+\frac{i}{2}\right) (a-i)\right) \\ -\text{Li}_2\left(\left(\frac{1}{2}-\frac{i}{2}\right) (a+i)\right) \\ +\text{Li}_2\left(\left(-\frac{1}{2}-\frac{i}{2}\right) (a+i)\right)\right) \end{align}$$

As the two lines marked $(*1*)$ and $(*2*)$ are already similar to the final result for $a=\sqrt {3}$ there should be a lot of cancellation from the polylog functions.

(to be completed)