My brother's friend gave me the following wicked integral with a beautiful result

\begin{equation} {\Large\int_0^\infty} \frac{dx}{\sqrt{x} \bigg[x^2+\left(1+2\sqrt{2}\right)x+1\bigg] \bigg[1-x+x^2-x^3+\cdots+x^{50}\bigg]}={\large\left(\sqrt{2}-1\right)\pi} \end{equation}

He claimed the above integral can be generalised to the following form \begin{equation} {\Large\int_0^\infty} \frac{dx}{\sqrt{x} \bigg[x^2+ax+1\bigg] \bigg[1-x+x^2-x^3+\cdots+(-x)^{n}\bigg]}=\ldots \end{equation} This is a challenging problem. How to prove it and what is the closed-form of the general integral?

  • 43
    $\begingroup$ Is your brother's friend named Cleo? $\endgroup$ – Git Gud Oct 17 '14 at 18:00
  • 2
    $\begingroup$ Let $x=-y$, then $1-x+x^2-x^3+\cdots+x^{50}=1+y+y^2+y^3+\cdots+y^{50}=\frac{1-y^{51}}{1-y}$ $\endgroup$ – mike Oct 17 '14 at 18:01
  • 3
    $\begingroup$ Hint: 1) substitute $x$ by $1/x$ and combine the new integral with the old to get rid of the horrible factor at end 2) change variable to $u = \sqrt{x} - \frac{1}{\sqrt{x}}$. $\endgroup$ – achille hui Oct 17 '14 at 18:04
  • 13
    $\begingroup$ It's a special case of the integral $$\int_{0}^{\infty} \frac{1}{\sqrt{x}} \left(\frac{x}{x^{2}+2ax+1} \right)^{r} \frac{x+1}{x(x^{s}+1)} \ dx = \frac{B(r - \frac{1}{2}, \frac{1}{2})}{2^{r-1/2} (1+a)^{r-1/2}} . $$ There is an entire chapter in the book Irresistible Integrals devoted to this integral. $\endgroup$ – Random Variable Oct 18 '14 at 3:37
  • 2
    $\begingroup$ It's a good book, but not so good that I would necessarily recommend buying it. That paper covers the most interesting chapter in the book. $\endgroup$ – Random Variable Oct 19 '14 at 15:05

Indeed let $$ I(n,a)=\int_0^\infty\frac{dx}{\sqrt{x}(1+ax+x^2)(\sum_{k=0}^n(-x)^k)} $$ The change of variables $x\leftarrow 1/x$ yields $$ I(n,a)=\int_0^\infty\frac{(-1)^nx^{n+1}dx}{ \sqrt{x}(1+ax+x^2)(\sum_{k=0}^n(-x)^k)} $$ Thus $$ 2I(n,a)=\int_0^\infty\frac{1+x}{\sqrt{x}(1+ax+x^2)}dx= 2\int_0^\infty\frac{1+t^2}{ 1+at^2+t^4}dt $$ Or equivalently, setting $u=t-1/t$, $$ I(n,a)= \int_{-\infty}^\infty\frac{du}{ 2+a+u^2} =\frac{\pi}{\sqrt{2+a}}. $$

  • 9
    $\begingroup$ I think it is very unintuitive that this is independent of $n$. It would be nice if someone offered an intuitive explanation of that fact. $\endgroup$ – becko Oct 17 '14 at 22:09
  • 1
    $\begingroup$ Nice answer. It's all true if $n \geq 0$ and $a > -2$. $\endgroup$ – user153012 Oct 17 '14 at 23:40
  • 6
    $\begingroup$ Very nice answer Prof! ٩(˘◡˘)۶ $\endgroup$ – Anastasiya-Romanova 秀 Oct 19 '14 at 12:49

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.