4
$\begingroup$

Again, improper integrals involving $\ln(1+x^2)$

I am trying to get a result for the integral $I_{\alpha}=\int_{0}^{\infty} \frac{\ln(1+x^2)}{x^\alpha}dx$ - asked above link- using some complex analysis, however, I couldn't find an appropriate solution. (Of course, for $\alpha \in (0,3)$ as stated above link (o.w. it is divergent))

Any ideas?

$\endgroup$
1
  • $\begingroup$ Applying by parts with $u=ln(1+x^2)$ and $v'=x^{- \alpha}$ we have: $\int uv'= x^{-1-\alpha} ln(1+x^2) - \int 2dx/((1+x^2)x^{\alpha})$ then I guess branch cut? $\endgroup$
    – user108539
    Nov 16 '14 at 19:52
5
$\begingroup$

Let $\alpha = 1 + p$. Then we have $$ \int_0^{\infty}\frac{\ln(1+x^2)}{x^{1+p}}dx $$ where $0<p<2$. Let $u = \ln(1+x^2)$ and $dv = x^{-1-p}dx$. Then $du = \frac{2x}{1+x^2}$ and $v = \frac{-1}{px^p}$. $$ uv\big|_0^{\infty}\to 0 $$ for $p\in(0,2)$. $$ \int_0^{\infty}\frac{\ln(1+x^2)}{x^{1+p}}dx=\frac{2}{p}\int_0^{\infty}\frac{x^{1-p}}{x^2+1}dx $$ Let $\beta -1 = 1-p$ so $p=2-\beta$. Let $x=t^2$ so $dx=2tdt$. $$ \frac{4}{2-\beta}\int_0^{\infty}\frac{t^{2\beta - 1}}{t^4+1}dt $$ The contour we desire is enter image description here

The integrals over the semicircles go to zero as $R\to\infty$ and $\delta\to 0$ so we have \begin{align} \frac{4}{2-\beta}\int_0^{\infty}\frac{t^{2\beta - 1}}{t^4+1}dt &= \frac{4}{2-\beta}\int_{-\infty}^{\infty}\frac{z^{2\beta - 1}}{z^4+1}dz\\ &= \frac{4}{2-\beta}\int_0^{\infty}\frac{z^{2\beta - 1}+(-z)^{2\beta - 1}}{z^4+1}dz \end{align} Now, $(-z)^{2\beta} = -e^{2\pi i\beta}z^{2\beta}$. $$ \frac{4}{2-\beta}\int_0^{\infty}\frac{z^{2\beta - 1}+(-z)^{2\beta - 1}}{z^4+1}dz = \frac{4(1-e^{2\pi i\beta})}{2-\beta}\int_0^{\infty}\frac{z^{2\beta - 1}}{z^4+1}dz\tag{1} $$ Then $(1)$ is equal to $2\pi i$ times sum of the residue is the upper half plane. \begin{align} \frac{8\pi i}{(1-e^{2\pi i\beta})(2-\beta)}\Bigl[\lim_{z\to e^{\frac{i \pi }{4}}}\frac{(z-e^{\frac{i \pi }{4}}) z^{2\beta -1}}{z^4+1}+\lim_{z\to e^{\frac{3i\pi}{4}}}\frac{(z-e^{\frac{3i\pi}{4}})z^{2\beta-1}}{z^4+1}\Bigr] &= \frac{\pi\csc\bigl(\frac{\pi\beta}{2}\bigr)}{2-\beta}\\ &= \frac{\pi\csc\bigl(\frac{\pi p}{2}\bigr)}{p}\tag{2}\\ &= \frac{\pi\sec\bigl(\frac{\pi\alpha}{2}\bigr)}{1-\alpha}\tag{3} \end{align} Equations $(2)$ and $(3)$ occur by back substituting for $\beta$ and $p$.

$\endgroup$
1
$\begingroup$

Differentiation under the integral sign is another possibility. Let: $$ I(\alpha,\beta) = \int_{0}^{+\infty}\frac{\log(1+\beta x^2)}{x^\alpha}\,dx. \tag{1}$$ Then assuming $1<\alpha<3$ we have: $$\frac{\partial}{\partial\beta} I(\alpha,\beta) = \int_{0}^{+\infty}\frac{x^{2-\alpha}}{1+\beta x^2}\,dx = \beta^{\frac{\alpha-3}{2}}\int_{0}^{+\infty}\frac{x^{2-\alpha}}{1+x^2}\,dx\tag{2}$$ and by replacing $x$ with $\tan\theta$ we get, through the Euler beta function and the reflection formula for the $\Gamma$ function: $$\frac{\partial}{\partial\beta} I(\alpha,\beta) = \beta^{\frac{\alpha-3}{2}}\int_{0}^{\pi/2}\left(\tan\theta\right)^{2-\alpha}\,d\theta = -\beta^{\frac{\alpha-3}{2}}\frac{\pi}{2\cos\left(\frac{\pi \alpha}{2}\right)}\tag{3}$$ and by integrating $(3)$ with respect to $\beta$ we have: $$ \int_{0}^{+\infty}\frac{\log(1+x^2)}{x^\alpha}\,dx = \frac{\pi}{1-\alpha}\,\sec\left(\frac{\pi\alpha}{2}\right).\tag{4}$$

$\endgroup$
1
  • $\begingroup$ @dustin: I did not feel insulted in any way, I was just working on a completely different problem. And upvoting is not compulsive for sure :) $\endgroup$ Mar 29 '15 at 21:38

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.