10

I was trying do variations of an integral representation for $\log\frac{\pi}{2}$ due to Jonathan Sondow, when I am wondering about if it is possible to evaluate

$$\int_0^1-\frac{1-x}{(1-x+x^2)\log x}\,dx,\tag{1}$$ Wolfram Alpha online calculator provide me a closed-form with code

int -(1-x)/((1-x+x^2)log(x)) dx, from x=0 to x=1

Question. Please provide me hints to know how evaluate previous this definite integral as $$\int_0^1-\frac{1-x}{(1-x+x^2)\log x}\,dx=\log \left(\frac{\Gamma(1/6)}{\Gamma(2/3)}\right)-\frac{\log \pi}{2}$$ as said Wolfram Alpha. Many thanks.

Alex Ortiz
  • 24,844
  • 1
    Maybe this might be helpful: $x^2-x+1=\frac{x^3+1}{x+1}$ – Simply Beautiful Art Jul 16 '17 at 12:29
  • Many thanks, then you are saying $\int_0^1\frac{x^2-1}{(x^3+1)\log x}dx$ @SimplyBeautifulArt –  Jul 16 '17 at 12:32
  • why not sharing the WA result with us? – tired Jul 16 '17 at 12:32
  • I've provided the code, but if you want I can edit the post. Any case my apologizes @tired –  Jul 16 '17 at 12:33
  • 2
    i think this would make the question much better... – tired Jul 16 '17 at 12:34
  • If there are some user that thinks that this question could be interesting, I tried also a variantion using the polynomial $1+x-x^2$ in the denominator, but I don't know a closed form for this different example. Many thanks all users. –  Jul 16 '17 at 12:38
  • Many thanks for the upvotes, I prefer in this question upvotes for the users who answered this question, that have the merit. –  Jul 16 '17 at 13:08
  • i messing up some algebra and haven't to much time today so here as an extended comment: We explore the class of paramter dependent integrals

    $$J(a)=\int_0^{\infty}dx\frac{e^{at}(e^{2x}-1)}{(e^{3x}+1) x}$$

    Now consider a derivative with respect to $a$ we get (after rescaling) $$ 3J'a=\int_0^{\infty}\frac{e^{ (a+2)x}-e^{a x}}{e^x+1}=I'{a/3+2/3}-I'_{a/3} $$

    expanding the denominator as geometric series and integrating termwise yields ...

    – tired Jul 16 '17 at 13:10
  • @user243301 :P Don't tell us how to vote. If we wanna upvote the question, we'll upvote the question – Simply Beautiful Art Jul 16 '17 at 13:10
  • 1
    $$2I'{a/3}=\sum{n\geq0}\frac{(-1)^n}{n+a/3}=\psi(1-\frac a6)-\psi(\frac12-\frac a6)$$

    and likewise $$ I'{a/3+2/3}=\sum{n\geq0}\frac{(-1)^n}{n+a/3}=\psi(\frac 13-\frac a6)-\psi(\frac16-\frac a6) $$

    where $\psi(z) =\log(\Gamma(z))'$ is the digamma function the digamma function. Integrating with respect to $a$ therefore yields

    $$6J_a=\log(\Gamma(1-\frac a6))-\log(\Gamma(\frac12-\frac a6))-\log(\Gamma(\frac13-\frac a6))+\log(\Gamma(\frac16-\frac a6))+C$$

    Now your integral in question is just $J_1$...

    – tired Jul 16 '17 at 13:11
  • 1
    @tired Or you just let $t=e^x$, and apply some complex analysis. – Simply Beautiful Art Jul 16 '17 at 13:11
  • @SimplyBeautifulArt what kind of complex analyis do you have in mind? Abel-Plana? – tired Jul 16 '17 at 13:12
  • Many thanks @tired now I am saturated, but surely that your comments are valuable and interesting for other users. I am going in next future read these. –  Jul 16 '17 at 13:12
  • Similarly, I am going to study your hints in next future. Many thanks @SimplyBeautifulArt –  Jul 16 '17 at 13:15
  • @tired Hm, it looks doable, but I too am too busy today to look into it further xD – Simply Beautiful Art Jul 16 '17 at 13:15
  • 1
    Please don't use int -(1-x)/((1-x+x^2)log(x)) dx, from x =0 to x=1 to replace $$\int_0^1 -\frac{(1-x)}{(1-x + x^2)\log(x)} = \int_0^1 \frac{x-1}{(x^2 -x+1)\log(x)}$$ – amWhy Jul 16 '17 at 16:13
  • Many thanks @amWhy my intention was provide the code implemented in Wolfram Language, and do a comparison with the identity due to Jonathan Sondow. Any case I accept your words. –  Jul 16 '17 at 20:07

6 Answers6

10

Thanks to the power series representation of $1/(1+x^3)$ and the Dominated Convergence Theorem, the given integral is \begin{align*} -\int_0^1\frac{1-x^2}{(1+x^3)\log x}\,dx &=\int_0^1\sum_{k=0}^{\infty}(-1)^{k+1}\frac{(1-x^2)x^{3k}}{\log x}\,dx\\ &=\int_0^1\sum_{k=0}^{\infty}\frac{(1-x^2)(x^{6k+3}-x^{6k})}{\log x}\,dx\\ &=\sum_{k=0}^{\infty}\int_0^1\frac{(1-x^2)(x^{6k+3}-x^{6k})}{\log x}\,dx\\&= \sum_{k=0}^{\infty}(\ln(6k+3)-\ln(6k+1)+\ln(6k+4)-\ln(6k+6))\\&= \ln\left(\prod_{k=0}^{\infty}\frac{(6k+3)(6k+4)}{(6k+1)(6k+6)}\right) =\ln\left(\frac{\Gamma(1/6)}{\sqrt{\pi}\Gamma(2/3)}\right) \end{align*} where $\Gamma(x)$ is the Gamma function and we used the fact that $$\int_{0}^{1}\frac{x^n-1}{\log x}\,dx = \int_{0}^{+\infty}\frac{1-e^{-nt}}{t}\,e^{-t}dt = \log(n+1)$$ (apply Frullani's integral in the last step).

Robert Z
  • 145,942
  • Hm, this probably results in some harmonic numbers. – Simply Beautiful Art Jul 16 '17 at 12:48
  • Many thanks also to you. –  Jul 16 '17 at 12:50
  • How did he interchange the sum and the integral that's what I want to know. – Zophikel Jul 16 '17 at 12:55
  • @Zophikel Probably DCT – Simply Beautiful Art Jul 16 '17 at 12:56
  • What's DCT if I may ask, also it should be trival to prove that:$$ -\int_0^1\frac{1-x^2}{(1+x^3)\log x},dx=\sum_{k=0}^{\infty}(-1)^{k+1}\int_0^1\frac{(1-x^2)x^{3k}}{\log x},dx\= \sum_{k=0}^{\infty}(-1)^{k+1}(\ln(3k+1)-\ln(3k+3))$$ Via Asymptotic Methods – Zophikel Jul 16 '17 at 12:57
  • @Zophikel the dominated convergence theorem, often used for interchanging integrals and sums. – Simply Beautiful Art Jul 16 '17 at 12:58
  • 1
    @Zophikel Yes to interchange the sum and the integral I used Dominated Convergence – Robert Z Jul 16 '17 at 12:59
  • Ahhhh, ok would it appropriate to post an answer verifying the hint :>) – Zophikel Jul 16 '17 at 12:59
  • 1
    Although I am certain that your computation can be justified, I would like to mention that the Fubini's theorem does not apply directly since $$ \int_{0}^{1} \sum_{k=0}^{\infty} \left| (-1)^{k+1} \frac{(1-x^2)x^{3k}}{\log x} \right| , dx = \infty. $$ You may need more elaborated technique to justify that step, for instance regularizing the denominator and using a combination of both Fubini's theorem and the dominated convergence theorem. – Sangchul Lee Jul 16 '17 at 13:00
  • 1
    Hmmmm,,, now looking back at the question maybe there's a simpler way via Asymptotics – Zophikel Jul 16 '17 at 13:03
  • @Sangchul Lee Now the problem of interchanging sum and integral should be solved. – Robert Z Jul 16 '17 at 13:14
  • Oh yes, it seems fine now. We can count on the Tonelli's theorem instead to interchange the sum and integral. (+1) – Sangchul Lee Jul 16 '17 at 13:17
  • ah nice..that is a bit simpler then my approach (+1) – tired Jul 16 '17 at 13:22
  • 2
    I also tried a similar approach as @tired. I computed the integral $$ I(s, a) := \int_{0}^{\infty} \frac{t^{s-1}e^{-at}}{1+e^{-t}} , dt = \frac{\Gamma(s)}{2^s} \left[ \zeta\left(s, \frac{a}{2}\right) - \zeta\left(s, \frac{a+1}{2}\right) \right] $$ and then tried to use a functional equation for Hurwitz zeta function to compute the limit $$ \int_{0}^{1} \frac{1-x}{(1-x+x^2)\log(1/x)} , dx = \lim_{s\to 0} [ I(s, \tfrac{1}{3}) - I(s, 1) ]. $$ But your approach is much neater and simpler. – Sangchul Lee Jul 16 '17 at 13:28
  • 1
    @SangchulLee yeah the limiting procedure becomes quite daunting :( – tired Jul 16 '17 at 13:29
9

I was confused by the implicit use of the Frullani integral, so I think it bears mention: $$ \begin{align} \int_0^1\frac{x^n-x^m}{\log(x)}\,\mathrm{d}x &=\int_0^\infty\frac{e^{-mu}-e^{-nu}}{u}e^{-u}\,\mathrm{d}u\\ &=\lim_{\epsilon\to0^+}\int_\epsilon^\infty\frac{e^{-(m+1)u}-e^{-(n+1)u}}{u}\,\mathrm{d}u\\ &=\lim_{\epsilon\to0^+}\int_{(m+1)\epsilon}^{(n+1)\epsilon}\frac{e^{-u}}{u}\,\mathrm{d}u\\[3pt] &=\log\left(\frac{n+1}{m+1}\right) \end{align} $$ $$ \begin{align} -\int_0^1\frac{1-x}{\left(1-x+x^2\right)\log(x)}\,\mathrm{d}x &=-\int_0^1\frac{1-x^2}{\left(1+x^3\right)\log(x)}\,\mathrm{d}x\\ &=-\sum_{k=0}^\infty(-1)^k\frac{x^{3k}-x^{3k+2}}{\log(x)}\,\mathrm{d}x\\ &=\sum_{k=0}^\infty(-1)^k\log\left(\frac{3k+3}{3k+1}\right)\\ &=\sum_{k=0}^\infty\log\left(\frac{(6k+3)(6k+4)}{(6k+1)(6k+6)}\right)\\ &=\log\left(\prod_{k=0}^\infty\frac{\left(k+\frac12\right)\left(k+\frac23\right)}{\left(k+\frac16\right)(k+1)}\right)\\ &=\lim_{n\to\infty}\log\left(\prod_{k=0}^{n-1}\frac{\color{#C00}{\left(k+\frac12\right)}\color{#090}{\left(k+\frac23\right)}}{\color{#00F}{\left(k+\frac16\right)}(k+1)}\right)\\ &=\log\left(\lim_{n\to\infty}\color{#C00}{\frac{\Gamma\left(n+\frac12\right)}{\Gamma\left(\frac12\right)}}\color{#090}{\frac{\Gamma\left(n+\frac23\right)}{\Gamma\left(\frac23\right)}}\color{#00F}{\frac{\Gamma\left(\frac16\right)}{\Gamma\left(n+\frac16\right)}}\frac{\Gamma(1)}{\Gamma(n+1)}\right)\\ &=\log\left(\frac1{\sqrt\pi}\frac{\Gamma\left(\frac16\right)}{\Gamma\left(\frac23\right)}\right)+\log\left(\lim_{n\to\infty}\frac{\Gamma\left(n+\frac12\right)\Gamma\left(n+\frac23\right)}{\Gamma\left(n+\frac16\right)\Gamma(n+1)}\right)\\ &=\log\left(\frac1{\sqrt\pi}\frac{\Gamma\left(\frac16\right)}{\Gamma\left(\frac23\right)}\right) \end{align} $$ The last step is by Gautschi's Inequality.

robjohn
  • 345,667
2

I thought it might be instructive to present an approach that does not rely on Frullani's integral.

To that end, we first note that $\frac{x-1}{\log(x)}=\int_0^1 x^s\,ds$.

Therefore, we can write

$$\begin{align} \int_0^1 \frac{1+x}{1+x^3}\,\frac{x-1}{\log(x)}\,dx&=\int_0^1 \left(\int_0^1 \frac{x^s+x^{s+1}}{1+x^3} \right)\,ds\\\\ &=\int_0^1 \sum_{n=0}^\infty (-1)^n\int_0^1 (x^{s+3n}+x^{s+3n+1})\,dx\,ds\\\\ &=\int_0^1 \sum_{n=0}^\infty (-1)^n \left(\frac{1}{s+3n+1}+\frac{1}{s+3n+2}\right)\,ds\\\\ &=\sum_{n=0}^\infty (-1)^n \log\left(\frac{3n+3}{3n+1}\right)\\\\ &=\sum_{n=0}^\infty \log\left(\frac{(6n+3)(6n+4)}{(6n+1)(6n+6)}\right)\tag 1\\\\ &=\log\left(\frac{\Gamma(1/6)}{\sqrt{\pi}\Gamma(2/3)}\right)\tag2 \end{align}$$

where in going from $(1)$ to $(2)$ we relied on the analysis posted in Rob's solution herein.

Mark Viola
  • 179,405
  • The last = sign alone is a huge jump to anybody asking this, proposed with no justification. – Did Jul 17 '17 at 16:25
  • @Did There were already two posted solutions with a way forward on that huge jump. I thought it redundant to duplicate. Would you suggest that I duplicate those or just reference them? – Mark Viola Jul 17 '17 at 20:00
  • The sentence you added, at least makes clear that there is a gap. – Did Jul 17 '17 at 21:11
  • @Did I noticed that the accepted answer had the same gap. – Mark Viola Jul 17 '17 at 21:15
  • This is true. No idea why the value of the infinite product should be so obvious to any user asking this question. – Did Jul 17 '17 at 21:26
2

$\newcommand{\bbx}[1]{\,\bbox[15px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$ \begin{align} &-\int_{0}^{1}{1 - x \over \pars{1 - x + x^{2}}\ln\pars{x}}\,\dd x = \int_{0}^{1}\ \overbrace{{1 + x \over 1 + x^{3}}} ^{\ds{1 \over 1 - x + x^{2}}}\,\,\, \overbrace{{x - 1 \over \ln\pars{x}}}^{\ds{\int_{0}^{1}x^{t}\,\dd t}}\ \,\dd x = \int_{0}^{1}\int_{0}^{1}{x^{t} + x^{t + 1} \over 1 + x^{3}}\,\dd x\,\dd t \\[5mm] & = \int_{0}^{1}\int_{0}^{1}{x^{t} + x^{t + 1} - x^{t + 3} - x^{t + 4} \over 1 - x^{6}}\,\dd x\,\dd t \\[5mm] & \stackrel{x^{6}\ \mapsto\ x}{=}\,\,\, {1 \over 6}\int_{0}^{1}\int_{0}^{1} {x^{t/6 - 5/6} + x^{t/6 - 2/3} - x^{t/6 - 1/3} - x^{t/6 - 1/6} \over 1 - x}\,\dd x\,\dd t \\[5mm] & = {1 \over 6}\int_{0}^{1}\!\!\pars{% -\int_{0}^{1}\!\!{1 - x^{t/6 - 5/6} \over 1 - x}\,\dd x - \int_{0}^{1}\!\!{1 - x^{t/6 - 2/3} \over 1 - x}\,\dd x + \int_{0}^{1}\!\!{1 - x^{t/6 - 1/3} \over 1 - x}\,\dd x + \int_{0}^{1}\!\!{1 - x^{t/6 - 1/6} \over 1 - x}\,\dd x} \\[5mm] & = {1 \over 6}\ \overbrace{\int_{0}^{1}\bracks{% -\Psi\pars{{t \over 6} + {1 \over 6}} - \Psi\pars{{t \over 6} + {1 \over 3}} + \Psi\pars{{t \over 6} + {2 \over 3}} + \Psi\pars{{t \over 6} + {5 \over 6}}}\dd t} ^{\ds{\Psi:\ Digamma\ Function}} \\[5mm] & = \left.\ln\pars{\Gamma\pars{t/6 + 2/3}\Gamma\pars{t/6 + 5/6} \over \Gamma\pars{t/6 + 1/6}\Gamma\pars{t/6 + 1/3}}\right\vert_{\ 0}^{\ 1}\qquad \pars{~\Gamma:\ Gamma\ Function.\ \Psi\pars{z} \stackrel{\mrm{def.}}{=} \totald{\ln\pars{\Gamma\pars{z}}}{z}~} \\[5mm] & = \ln\pars{{\Gamma\pars{5/6}\Gamma\pars{1} \over \Gamma\pars{1/3}\Gamma\pars{1/2}}\, {\Gamma\pars{1/6}\Gamma\pars{1/3} \over \Gamma\pars{2/3}\Gamma\pars{5/6}}} \\[5mm] & = \ln\pars{{1 \over \Gamma\pars{1/2}}\, {\Gamma\pars{1/6} \over \Gamma\pars{2/3}}}\qquad \pars{\ds{\mbox{Note that}\ \Gamma\pars{1} = 1\ \mbox{and}\ \Gamma\pars{1 \over 2} = \root{\pi}}} \\[5mm] & = \bbx{\ln\pars{{1 \over \root{\pi}}\, {\Gamma\pars{1/6} \over \Gamma\pars{2/3}}}} \approx 0.8412 \end{align}

Note that $\ds{\left.\vphantom{\Large A}\Psi\pars{z}\right\vert_{\ \Re\pars{z}\ >\ 0} = -\gamma + \int_{0}^{1}{1 - t^{z - 1} \over 1 - t}\,\dd t}$. $\ds{\gamma}$ is the Euler-Mascheroni Constant. See $\ds{\color{#000}{\mathbf{6.3.22}}}$ in A & S Table.

Felix Marin
  • 89,464
0

Let

$$I=\int_0^1\frac{1+x}{1+x^3}\frac{x-1}{\ln(x)}dx$$ and

$$I(a)=\int_0^1\frac{1+x}{1+x^3}\frac{x^a-1}{\ln(x)}dx$$

and notice that $I(1)=I$ and $I(0)=0.$

$$I'(a)=\int_0^1\frac{x^a+x^{a+1}}{1+x^3}dx\overset{x=t^{1/3}}{=}\frac13\int_0^1\frac{t^{\frac{a-2}{3}}+t^{\frac{a-1}{3}}}{1+t}dt$$

By using

$$\int_0^1\frac{x^n}{1+x}dx=\frac12\psi\left(\frac{n+2}{2}\right)-\frac12\psi\left(\frac{n+1}{2}\right),$$

we have

$$I'(a)=\frac16\left[\psi\left(\frac{a+4}{6}\right)-\psi\left(\frac{a+1}{6}\right)+\psi\left(\frac{a+5}{6}\right)-\psi\left(\frac{a+2}{6}\right)\right].$$

Integrate both sides from $a=0$ to $a=1$

$$\int_0^1 I'(a)da=I(a)|_0^1=I(1)-I(0)=I-0$$

$$=\int_0^1 \frac16\left[\psi\left(\frac{a+4}{6}\right)-\psi\left(\frac{a+1}{6}\right)+\psi\left(\frac{a+5}{6}\right)-\psi\left(\frac{a+2}{6}\right)\right]da$$

$$=\left[\ln\Gamma\left(\frac{a+4}{6}\right)-\ln\Gamma\left(\frac{a+1}{6}\right)+\ln\Gamma\left(\frac{a+5}{6}\right)-\ln\Gamma\left(\frac{a+2}{6}\right)\right]_0^1$$

$$=\left[\ln\Gamma\left(\frac{5}{6}\right)-\ln\Gamma\left(\frac{1}{3}\right)+\ln\Gamma\left(1\right)-\ln\Gamma\left(\frac{1}{2}\right)\right]$$

$$-\left[\ln\Gamma\left(\frac{2}{3}\right)-\ln\Gamma\left(\frac{1}{6}\right)+\ln\Gamma\left(\frac{5}{6}\right)-\ln\Gamma\left(\frac{1}{3}\right)\right]$$

$$=-\ln\Gamma\left(\frac{1}{2}\right)-\ln\Gamma\left(\frac{2}{3}\right)+\ln\Gamma\left(\frac{1}{6}\right)$$

$$=\ln\left(\frac{\Gamma\left(\frac16\right)}{\sqrt{\pi}\, \Gamma\left(\frac23\right)}\right).$$

Ali Shadhar
  • 25,498
0

For simplicity, we consider the integral as $$ I=\int_0^1 \frac{x-1}{\left(1-x+x^2\right) \ln x} d x=\int_0^1 \frac{x^2-1}{\left(x^3+1\right) \ln x} d x $$ Considering the parametrised integral $$ I(a)=\int_0^1 \frac{x^a-1}{\left(x^3+1\right) \ln x} d x $$ Differentiating both sides w.r.t. $a$ gives $$ I^{\prime}(a)=\int_0^1 \frac{x^a}{x^3+1} d x=\frac{1}{6}\left[\psi\left(\frac{a+4}{6}\right)-\psi\left(\frac{a+1}{6}\right)\right] $$ Integrating $I^{\prime}(a)$ from $a=0$ to $a=2$ brings us $$ \begin{aligned} I & =I(2)-I(0)=\frac{1}{6} \int_0^2\left[\psi\left(\frac{a+4}{6}\right)-\psi\left(\frac{a+1}{6}\right)\right] d a \\ & =\left[\ln \Gamma\left(\frac{a+4}{6}\right)-\ln \Gamma\left(\frac{a+1}{6}\right)\right]_0^2=-\ln \Gamma\left(\frac{1}{2}\right)-\ln \Gamma\left(\frac{2}{3}\right)+\ln \Gamma \left(\frac{1}{6}\right)\\&= \ln \left(\frac{\Gamma\left(\frac{1}{6}\right)}{\sqrt{\pi} \Gamma\left(\frac{2}{3}\right)}\right) \end{aligned} $$

Lai
  • 20,421