For what $\alpha\in\mathbb{R}$ does $$\int_{1}^{\infty}\frac{\ln(1+x^{\alpha})}{\sqrt{x^2-1}} \ dx$$ converge?

I know this is a duplicate problem, but I don't follow the accepted answer, in particular the explanation of case 3. I understand that we should look at the cases $\alpha\in(-\infty,0), \alpha\in(0,\infty)$ and $\alpha = 0$. My questions are:

  1. In this thread, Oliver Oloa compares, in case 1 & 2, his two simplified integrals with the integral $$\int_{b}^{\infty}\frac{1}{x} \ dx, \quad (b>0).$$ But why did he cho0se the restriction $b>0$ and not $b\geq 0$? I understand that $b$ can't be negative, but surely $\int_{0}^{\infty}\frac{1}{x} \ dx$ IS divergent as well.

  2. In case 3, he goes $$ \ln(1+x^\alpha) \sim \frac1{x^{|\alpha|}} $$ $$ \frac{\ln(1+x^\alpha)}{\sqrt{x^2-1}} \sim \frac1{x^{|\alpha|+1}} $$ I really need a breakdown of the steps in the entire case 3. How did the absolute values appear?

  3. Is there another way to solve this problem?

Note: Taylor expansions/L'Hopital's are not allowed to be used.

  • $\begingroup$ Are you allowed to use l'Hospital rule ? $\endgroup$ – Noether Nov 17 '17 at 11:25
  • $\begingroup$ No unfortunately :( $\endgroup$ – Parseval Nov 17 '17 at 14:18

Regarding case 3, you have that for negative negative $\alpha$

$$x^\alpha = \frac{1}{x^{|\alpha|}}$$

Since you are integrating from $1$ onwards, the argument of the logarithm is close to $1$ as $1/x^{|\alpha|}$ is close to $0$. Taylor's expansion of $\log(1+t)$ near $t=0$ tells you that

$$\log(1+t) \approx t$$

On the other hand, as $x$ grows bigger $\sqrt{x^2 -1} \approx \sqrt{x^2} = x$.


$$\frac{\log\left(1+\frac{1}{x^{|\alpha|}}\right)}{\sqrt{x^2 -1}} \approx \frac{1}{x^{|\alpha|+1}}$$

  • $\begingroup$ Hi, thanks for your post but we are not allowed to use Taylor expansions since whe haven't covered it in this course. $\endgroup$ – Parseval Nov 17 '17 at 9:14

If $\alpha\geq 0$ the given integral is greater than $\int_{1}^{+\infty}\frac{\log 2}{\sqrt{x^2-1}}\,dx$, which is divergent. If $\alpha<0$, by setting $\beta=-\alpha$ we have $\beta>0$ and we may prove that $$ \int_{1}^{+\infty}\frac{\log\left(1+\frac{1}{x^\beta}\right)}{\sqrt{x^2-1}}\,dx\leq\int_{1}^{+\infty}\frac{dx}{x^\beta\sqrt{x^2-1}} $$ is convergent by enforcing the substitution $x=\cosh t$, for instance, turnig the last integral into $$ \int_{0}^{+\infty}\frac{dt}{(\cosh t)^{\beta}}\leq \int_{0}^{+\infty}\frac{2^\beta}{e^{\beta t}}\,dt=\frac{2^\beta}{\beta}<+\infty.$$


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.