27
$\begingroup$

Today I discussed the following integral in the chat room

$$\int_0^\infty \ln \left( \frac{x^2+2kx\cos b+k^2}{x^2+2kx\cos a+k^2}\right) \;\frac{\mathrm dx}{x}$$ where $0\leq a, b\leq \pi$ and $k>0$.

Some users suggested me that I can use Frullani's theorem:

$$\int_0^\infty \frac{f(ax) - f(bx)}{x} = \big[f(0) - f(\infty)\big]\ln \left(\frac ab\right)$$ So I tried to work with that way. \begin{align} I&=\int_0^\infty \ln \left( \frac{x^2+2kx\cos b+k^2}{x^2+2kx\cos a+k^2}\right) \;\frac{\mathrm dx}{x}\\ &=\int_0^\infty \frac{\ln \left( x^2+2kx\cos b+k^2\right)-\ln \left( x^2+2kx\cos a+k^2\right)}{x}\mathrm dx\tag{1}\\ &=\int_0^\infty \frac{\ln \left( 1+\dfrac{2k\cos b}{x}+\dfrac{k^2}{x^2}\right)-\ln \left( 1+\dfrac{2k\cos a}{x}+\dfrac{k^2}{x^2}\right)}{x}\mathrm dx\tag{2}\\ \end{align} The issue arose from $(1)$ because $f(\infty)$ diverges and the same issue arose from $(2)$ because $f(0)$ diverges. I then tried to use D.U.I.S. by differentiating w.r.t. $k$, but it seemed no hope because WolframAlpha gave me this horrible form. Any idea? Any help would be appreciated. Thanks in advance.

$\endgroup$
28
$\begingroup$

First note that by substituting $x\mapsto kx$, we get $$ \int_0^\infty\log\left(\frac{x^2+2kx\cos(a)+k^2}{x^2+2kx\cos(b)+k^2}\right)\frac{\mathrm{d}x}{x} =\int_0^\infty\log\left(\frac{x^2+2x\cos(a)+1}{x^2+2x\cos(b)+1}\right)\frac{\mathrm{d}x}{x} $$ Let $u=\frac{x+\cos(a)}{\sin(a)}$. Then $$ \begin{align} \frac{\mathrm{d}}{\mathrm{d}a}\int_0^\infty\log\left(\frac{x^2+2x\cos(a)+1}{x^2+2x+1}\right)\frac{\mathrm{d}x}{x} &=-2\int_0^\infty\frac{\sin(a)}{x^2+2x\cos(a)+1}\,\mathrm{d}x\\ &=-2\int_0^\infty\frac{\sin(a)}{(x+\cos(a))^2+\sin^2(a)}\,\mathrm{d}x\\ &=-\frac2{\sin(a)}\int_0^\infty\frac1{\frac{(x+\cos(a))^2}{\sin^2(a)}+1}\,\mathrm{d}x\\ &=-2\int_{\cot(a)}^\infty\frac1{u^2+1}\,\mathrm{d}u\\[9pt] &=-2a \end{align} $$ Integrating in $a$ gives $$ \int_0^\infty\log\left(\frac{x^2+2x\cos(a)+1}{x^2+2x+1}\right)\frac{\mathrm{d}x}{x} =-a^2 $$ Therefore, by subtraction, $$ \int_0^\infty\log\left(\frac{x^2+2kx\cos(a)+k^2}{x^2+2kx\cos(b)+k^2}\right)\frac{\mathrm{d}x}{x} =b^2-a^2 $$

$\endgroup$
  • 1
    $\begingroup$ Hhmmm, your answer gives me a new idea to evaluate this integral. (+1) $\endgroup$ – Venus Dec 15 '14 at 17:14
  • $\begingroup$ Brilliant ! :-$)$ $\endgroup$ – Lucian Dec 15 '14 at 21:14
13
$\begingroup$

\begin{align} \int_0^\infty \ln \left( \frac{x^2+2kx\cos b+k^2}{x^2+2kx\cos a+k^2}\right) \;\frac{\mathrm dx}{x}&=\int_0^\infty\frac{1}{x}\int_a^b \frac{\mathrm d}{\mathrm dy}\ln \left(x^2+2kx\cos y+k^2\right) \;\mathrm dy\,\mathrm dx\\[10pt] &=-\int_0^\infty\frac{1}{x}\int_a^b \frac{2kx\sin y}{x^2+2kx\cos y+k^2} \;\mathrm dy\,\mathrm dx\\[10pt] &=2k\int_b^a\sin y\int_0^\infty \frac{\mathrm dx}{x^2+2kx\cos y+k^2} \;\mathrm dy\\[10pt] &=2k\int_b^a\sin y\int_0^\infty \frac{\mathrm dx}{\left(x+k\cos y\right)^2+k^2-k^2\cos^2y} \;\mathrm dy\\[10pt] &=2k\int_b^a\sin y\underbrace{\int_0^\infty \frac{\mathrm dx}{\left(x+k\cos y\right)^2+k^2\sin^2y}}_{\large\color{blue}{(k\sin y)\, u\,=\,x+k\cos y}} \;\mathrm dy\\[10pt] &=2\int_b^a\int_{\cot y}^\infty \frac{\mathrm du}{u^2+1} \;\mathrm dy\\[10pt] &=2\int_b^a\left(\frac{\pi}{2}-\arctan\left(\cot y\right)\right)\;\mathrm dy\\[10pt] &=2\int_b^ay\;\mathrm dy\\[10pt] &=\bbox[8pt,border:3px #FF69B4 solid]{\color{red}{\large a^2-b^2}} \end{align}

$\endgroup$
  • $\begingroup$ I see that you've encapsulated the differentiation in a double integral instead of integrating twice. $\endgroup$ – robjohn Dec 15 '14 at 17:49
  • $\begingroup$ @robjohn Yes. So, I didn't need to work twice. But my answer is inspired by your answer. Thanks. ^^ $\endgroup$ – Venus Dec 15 '14 at 17:51
11
$\begingroup$

Another approach is to use the Fourier series $$\sum_{k=1}^{\infty}\frac{x^{k} \cos(ka)}{k} = - \frac{1}{2} \log \left(x^{2} - 2 x \cos(a) +1 \right) \ , \ |x| <1 $$ which can be derived from the Maclaurin series of $\log(1-z)$ by replacing $z$ with $xe^{ia}$ and equating the real parts on both sides.

$$ \begin{align} \int_0^\infty\log\left(\frac{x^2+2kx\cos(b)+k^2}{x^2+2kx\cos(a)+k^2}\right)\frac{dx}{x} &=\int_0^\infty\log\left(\frac{u^2+2u\cos(b)+1}{u^2+2u\cos(a)+1}\right)\frac{du}{u} \tag{1} \\ &= 2 \int_{0}^{1} \tag{2} \log\left(\frac{u^2+2u\cos(b)+1}{u^2+2u\cos(a)+1}\right)\frac{du}{u} \\ &= 4 \int_{0}^{1} \frac{1}{u} \sum_{k=1}^{\infty} (-u)^{k} \frac{\cos(ka) - \cos(kb)}{k} \\ &= 4 \sum_{k=1}^{\infty} (-1)^{k} \frac{\cos(ka) -\cos(kb)}{k} \int_{0}^{1} u^{k-1} \ du \\ &= 4 \sum_{k=1}^{\infty} (-1)^{k} \frac{\cos(ka) - \cos(kb)}{k^{2}} \\ &= 4 \left(\frac{a^{2}}{4} - \frac{b^{2}}{4} \right) \tag{3} \\ &= a^{2}- b^{2} \end{align} $$

$ $

$(1)$ Let $ \displaystyle u= \frac{x}{k}.$

$(2)$ Separate the integral into two integrals, namely one over the interval $(0,1)$ and one over the interval $(1, \infty)$, and replace $u$ with $\frac{1}{u}$ in the second integral.

$(3)$ By integrating the Fourier series $ \displaystyle \sum_{k=1}^{\infty} (-1)^{k} \frac{\sin(ka)}{k} = - \frac{a}{2} \ , \ |a| < \pi$, we get $$ \sum_{k=1}^{\infty} (-1)^{k} \frac{\cos(ka)}{k^{2}} = \frac{a^{2}}{4} - \frac{\pi^{2}}{12} \ , \ |a| < \pi .$$

$\endgroup$
  • $\begingroup$ Very nice approach, (+1). Thanks for giving the nice Fouries series. $\endgroup$ – Venus Dec 16 '14 at 5:29
  • $\begingroup$ @Venus Thanks. I had accidentally put the parameter $a$ in the numerator and the parameter $b$ in the denominator. $\endgroup$ – Random Variable Dec 16 '14 at 8:39
  • $\begingroup$ How do you prove that Fourier series? $\endgroup$ – clathratus Feb 13 at 1:06
10
$\begingroup$

Here is a complex-analytic method: Notice that

$$ \int_{0}^{\infty} \log\left( \frac{x^{2} + 2x\cos b + 1}{x^{2} + 2x\cos a + 1} \right) \frac{dx}{x} = 2 \int_{0}^{\infty} \Re \frac{\log(1 + e^{ib}x) - \log(1 + e^{ia}x)}{x} \, dx. \tag{1} $$

Let $R$ be a positive large number. Then the function $z \mapsto \log(1+z)/z$ is analytic on $\Bbb{C} \setminus (-\infty, 1]$ with the standard branch cut and we get

\begin{align*} \int_{0}^{R} \frac{\log(1 + e^{ib}x)}{x} \, dx &=\int_{0}^{Re^{ib}} \frac{\log(1 + z)}{z} \, dz \qquad (z = e^{ib}x) \\ &= \int_{0}^{R} \frac{\log(1 + z)}{z} \, dz + \int_{R}^{Re^{ib}} \frac{\log(1 + z)}{z} \, dz \\ &= \int_{0}^{R} \frac{\log(1 + z)}{z} \, dz + i \int_{0}^{b} \log(1 + Re^{i\theta}) \, d\theta, \quad (z=Re^{i\theta}) \end{align*}

and likewise for the integral with $b$ replaced by $a$. This shows that

$$ \int_{0}^{R} \frac{\log(1 + e^{ib}x) - \log(1 + e^{ia}x)}{x} \, dx = i \int_{a}^{b} \log(1 + Re^{i\theta}) \, d\theta. $$

Multiplying by 2 and taking real part, we get

$$ 2 \int_{0}^{R} \Re \frac{\log(1 + e^{ib}x) - \log(1 + e^{ia}x)}{x} \, dx = - 2 \Im \int_{a}^{b} \log(R^{-1} + e^{i\theta}) \, d\theta. $$

(Here we exploited the fact that $\log R$ is real.) Taking $R \to \infty$, in view of the identity $\text{(1)}$ it follows that

$$ \int_{0}^{\infty} \log\left( \frac{x^{2} + 2x\cos b + 1}{x^{2} + 2x\cos a + 1} \right) \frac{dx}{x} = - \int_{a}^{b} 2\theta \, d\theta = a^{2} - b^{2}. $$

$\endgroup$
  • $\begingroup$ Thank you very much for helping me again, (+1). Anyway, Hatsune Miku looks very kawaii with that hat ^^ $\endgroup$ – Venus Dec 16 '14 at 5:26
  • 1
    $\begingroup$ @Venus, thank you! I tried many ways until the hat fits her perfectly. Your compliment rewards my effort :) $\endgroup$ – Sangchul Lee Dec 16 '14 at 6:41
7
$\begingroup$

$\newcommand{\angles}[1]{\left\langle\, #1 \,\right\rangle} \newcommand{\braces}[1]{\left\lbrace\, #1 \,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\, #1 \,\right\rbrack} \newcommand{\ceil}[1]{\,\left\lceil\, #1 \,\right\rceil\,} \newcommand{\dd}{{\rm d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\dsc}[1]{\displaystyle{\color{red}{#1}}} \newcommand{\expo}[1]{\,{\rm e}^{#1}\,} \newcommand{\fermi}{\,{\rm f}} \newcommand{\floor}[1]{\,\left\lfloor #1 \right\rfloor\,} \newcommand{\half}{{1 \over 2}} \newcommand{\ic}{{\rm i}} \newcommand{\iff}{\Longleftrightarrow} \newcommand{\imp}{\Longrightarrow} \newcommand{\Li}[1]{\,{\rm Li}_{#1}} \newcommand{\norm}[1]{\left\vert\left\vert\, #1\,\right\vert\right\vert} \newcommand{\pars}[1]{\left(\, #1 \,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\pp}{{\cal P}} \newcommand{\root}[2][]{\,\sqrt[#1]{\vphantom{\large A}\,#2\,}\,} \newcommand{\sech}{\,{\rm sech}} \newcommand{\sgn}{\,{\rm sgn}} \newcommand{\totald}[3][]{\frac{{\rm d}^{#1} #2}{{\rm d} #3^{#1}}} \newcommand{\ul}[1]{\underline{#1}} \newcommand{\verts}[1]{\left\vert\, #1 \,\right\vert}$ $\ds{\int_{0}^{\infty}\ln\pars{x^{2} + 2kx\cos\pars{b} + k^{2}\over x^{2} + 2kx\cos\pars{a} + k^2}\,{\dd x \over x}:\ {\large ?} \,,\qquad 0 <\ a\,,\ b\ <\pi\,,\quad k > 0.}$

\begin{align}&\overbrace{\color{#66f}{\large\int_{0}^{\infty} \ln\pars{x^{2} + 2kx\cos\pars{b} + k^{2}\over x^{2} + 2kx\cos\pars{a} + k^2}\,{\dd x \over x}}} ^{\dsc{x \over k}\ \ds{\mapsto}\ \dsc{x}} \\[5mm]&=\int_{0}^{\infty} \ln\pars{x^{2} + 2x\cos\pars{b} + 1\over x^{2} + 2x\cos\pars{a} + 1} \,{\dd x \over x} =\lim_{R\ \to\ \infty}\bracks{\fermi\pars{a,R} - \fermi\pars{b,R}} \\[5mm]&\mbox{where}\ \begin{array}{|c|}\hline\\ \ \fermi\pars{\mu,R}\equiv \int_{0}^{R}\ln\pars{x} {2x + 2\cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x\,,\quad 0 < \mu < \pi\ \\ \\ \hline \end{array}\qquad\qquad\quad\pars{1} \end{align}

Note that $\ds{r_{-} \equiv \exp\pars{\bracks{\pi - \mu}\ic}}$ and $\ds{r_{+} \equiv \exp\pars{\bracks{\pi + \mu}\ic}}$ are the roots of $\ds{x^{2} + 2x\cos\pars{\mu} + 1 = 0}$ such that $\ds{\pars{~\mbox{with}\ R > 1~}}$:

\begin{align}&\dsc{\int_{0}^{R}\ln^{2}\pars{x} {x + \cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x} \\[5mm]&=2\pi\ic\bracks{\half\,\ln^{2}\pars{r_{-}} + \half\,\ln^{2}\pars{r_{+}}} -\int_{R}^{0}\bracks{\ln\pars{x} + 2\pi\ic}^{2} {x + \cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x \\[5mm]&=\pi\ic\bracks{-\pars{\pi - \mu}^{2} - \pars{\pi + \mu}^{2}} +\dsc{\int_{0}^{R}\ln^{2}\pars{x} {x + \cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x} \\[5mm]&+2\pi\ic\ \overbrace{\int_{0}^{R}\ln\pars{x} {2x + 2\cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x} ^{\dsc{\fermi\pars{\mu,R}}}\ +\ \pars{2\pi\ic}^{2}\int_{0}^{R} {x + \cos\pars{\mu} \over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x \\[5mm]&-{\mathfrak C}\pars{R,\mu} \end{align} where $\ds{\left.{\mathfrak C}\pars{R,\mu} \equiv\oint\ln^{2}\pars{z} {z + \cos\pars{\mu}\over z^{2} + 2z\cos\pars{\mu} + 1}\,\dd z\,\right\vert _{z\ \equiv\ R\expo{\ic\theta}\,,\ 0\ <\ \theta\ <\ 2\pi}}$

This expression leads to: \begin{align} 0&=-2\pi\ic\pars{\pi^{2} + \mu^{2}} + 2\pi\ic\int_{0}^{R}\ln\pars{x} {2x + 2\cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x \\[5mm]&+\pars{2\pi\ic}^{2}\int_{\cos\pars{\mu}}^{R + \cos\pars{\mu}} {x \over x^{2} + \sin^{2}\pars{\mu}}\,\dd x - {\mathfrak C}\pars{R,\mu} \end{align}

and \begin{align} \fermi\pars{\mu,R}&=\int_{0}^{R}\ln\pars{x} {2x + 2\cos\pars{\mu}\over x^{2} + 2x\cos\pars{\mu} + 1}\,\dd x \\[5mm]&=\pi^{2} + \mu^{2} -\pi\ic\ln\pars{\bracks{R + \cos\pars{\mu}}^{2} + \sin^{2}\pars{\mu}} + {{\mathfrak C}\pars{R,\mu} \over 2\pi\ic} \end{align}

In the $\ds{R \to \infty}$ we'll have: \begin{align} \lim_{R \to\ \infty}\bracks{\fermi\pars{a,R} - \fermi\pars{b,R}} &=a^{2} - b^{2} \end{align}

such that expression $\pars{1}$ is reduced to: \begin{align}&\color{#66f}{\large\int_{0}^{\infty} \ln\pars{x^{2} + 2kx\cos\pars{b} + k^{2}\over x^{2} + 2kx\cos\pars{a} + k^2}\,{\dd x \over x}} =\color{#66f}{\large a^{2} - b^{2}} \end{align}

$\endgroup$
  • $\begingroup$ Thanks for your answer Felix, (+1). It's good to have many different ways to evaluate this integral. $\endgroup$ – Venus Dec 16 '14 at 6:46
  • $\begingroup$ @Venus You're welcome. I left some details that you can recover. I did it because it was growing out of my energy. $\endgroup$ – Felix Marin Dec 16 '14 at 7:10
  • $\begingroup$ Thanks for the details Felix, you're so kind. I hope I can learn some complex analysis techniques from your answer. Anyway, where's your hat? You should wear it. :-) $\endgroup$ – Venus Dec 16 '14 at 8:19

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.