$\phi$; golden ratio

$$I=\int_{0}^{\infty}{1\over (1+{\phi^{-2}}x^2)(1+{\phi^{-4}}x^2)}dx={\pi\over 2}\tag1$$

Let $q=\phi^{-2}$ and $p=\phi^{-4}$

$${1\over (1+qx^2)(1+px^2)}={Ax+B\over 1+qx^2}+{Cx+D\over 1+px^2}\tag2$$


$A=0$, $B=\phi$, $C=0$ and $D=-{1\over \phi}$

Sub the result into $(2)$ and we got

$${1\over (1+\phi^{-2}x^2)(1+\phi^{-4}x^2)}={\phi\over 1+\phi^{-2}x^2}-{\phi^{-1}\over 1+\phi^{-4}x^2}$$


$$I=\int_{0}^{\infty}\left({\phi\over 1+\phi^{-2}x^2}-{\phi^{-1}\over 1+\phi^{-4}x^2}\right)dx\tag3$$

Simplified $(3)\rightarrow (4)$

$$I=\phi^3\int_{0}^{\infty}{1\over \phi^2+x^2}dx-\phi^3\int_{0}^{\infty}{1\over \phi^4+x^2}dx\tag4$$


$$\int_{0}^{\infty}{1\over a^2+x^2}dx={\pi \over 2a}\tag5$$

Utilise $(5)$ to brings $(4)$ to $(6)$

$$I=\phi^3\cdot{\pi\over 2\phi}-\phi^3{\pi\over 2\phi^2}\tag6$$

$$I={\pi\over 2}(\phi^2-\phi)={\pi\over 2}\tag7$$

This method it is a bit tedious, can someone tackle integral (1) with a quicker method?

  • $\begingroup$ Partial Fraction is the algorithm that one may use to tackle this integral. Off the top of my head, I don't think there would be easier methods for this case. :) $\endgroup$ – H. R. Jun 12 '16 at 11:51
  • 1
    $\begingroup$ Perhaps an application of the Parseval theorem, but I don't know if that could be considered to be easier or quicker. $\endgroup$ – mickep Jun 12 '16 at 11:55
  • $\begingroup$ Personally, this is the fastest way I would solve this integral: By using partial fraction decomposition and then using the $\frac{1}{a^2+x^2}$ integral identity. It's a bit tedious, but integrals with quadratics in the denominator are always tedious in my opinion. $\endgroup$ – Noble Mushtak Jun 12 '16 at 12:02
  • $\begingroup$ A small note: In doing partial fraction decomposition, you don't need to have $A$ and $C$ in your ansatz, since your original function is even. Once that is noticed, the partial fraction decomposition can be done in your head. Now, I see no good reason (except, perhaps, learning new techniques) to use some other method. $\endgroup$ – mickep Jun 12 '16 at 18:26
  • $\begingroup$ You could simplify the path as mickep mentioned and also note that $$\phi^2(1+\phi^{-4}x^2)-(1+\phi^{-2}x^2)=\phi^2-1$$ $\endgroup$ – Math-fun Jun 13 '16 at 11:48

Through the residue theorem, by setting $f(x)=\frac{1}{\left(1+\frac{x^2}{\varphi^2}\right)\left(1+\frac{x^2}{\varphi^4}\right)}$ we have:

$$\begin{eqnarray*} I = \pi i \cdot\!\!\!\!\!\! \sum_{z\in\left\{i\varphi,i\varphi^2\right\}}\!\!\!\text{Res}(f(x),x=z)&=&\pi i\left(\frac{i\varphi^5}{2(\varphi^2-\varphi^4)}+\frac{i\varphi^4}{2(\varphi^4-\varphi^2)}\right)\\&=&\frac{\pi}{2}\cdot(\varphi^2-\varphi)\cdot\frac{\varphi^3}{\varphi^4-\varphi^2}\\&=&\frac{\pi}{2}\cdot 1\cdot 1 = \color{red}{\frac{\pi}{2}}.\end{eqnarray*} $$

An alternative is to prove a more general statement: $$ \forall A,B>0,\qquad I(A,B)=\int_{0}^{+\infty}\frac{dx}{\left(1+\frac{x^2}{A^2}\right)\left(1+\frac{x^2}{B^2}\right)}=\frac{\pi}{2}\cdot\frac{AB}{A+B}$$ through Lagrange's identity and Glasser's master theorem, or the fact that the Fourier transform of a Cauchy distribution is a Laplace distribution. For instance, Lagrange and Glasser give:

$$\begin{eqnarray*}I(A,B)=\int_{0}^{+\infty}\frac{dx}{\left(\frac{A+B}{AB}x\right)^2+\left(1-\frac{x^2}{AB}\right)^2}&=&\frac{AB}{A+B}\int_{0}^{+\infty}\frac{dx}{x^2+\left(1-\frac{AB x^2}{(A+B)^2}\right)^2}\end{eqnarray*}$$

where $\int_{0}^{+\infty}\frac{dx}{x^2+(1-k^2 x^2)^2}$ constantly equals $\frac{\pi}{2}$ for any $k\in\mathbb{R}\setminus\{0\}$.


I think a quite simple approach is with the residue theorem. We have $$\begin{align} \int_{0}^{\infty}\frac{1}{\left(1+\left(\frac{x}{\phi}\right)^{2}\right)\left(1+\left(\frac{x}{\phi^{2}}\right)^{2}\right)}dx= & \phi^{2}\int_{0}^{\infty}\frac{1}{\left(1+\phi^{2}x^{2}\right)\left(1+x^{2}\right)}dx \\ = & \phi^{2}\int_{0}^{\infty}f\left(x\right)dx \end{align} $$ so if we take as contour the semicircle with $\textrm{Im}\left(x\right)>0 $ and with the diameter on the real axis from $-R $ to $R $. It is not difficult to see that the integral on the semicircle go to zero if $R\rightarrow\infty $. So $$\begin{align} \int_{-\infty}^{\infty}\frac{1}{\left(1+\phi^{2}x^{2}\right)\left(1+x^{2}\right)}dx= &2\pi i\left(\textrm{Res}_{x=i}f\left(x\right)+\textrm{Res}_{x=\frac{i}{\phi}}f\left(x\right)\right) \\ = & \pi\left(\frac{1}{1-\phi^{2}}+\frac{\phi}{\phi^{2}-1}\right) \end{align} $$ then, by symmetry, $$\phi^{2}\int_{0}^{\infty}\frac{1}{\left(1+\phi^{2}x^{2}\right)\left(1+x^{2}\right)}dx=\frac{\pi}{2}\left(\frac{\phi^{2}}{1-\phi^{2}}+\frac{\phi^{3}}{\phi^{2}-1}\right)=\frac{\pi}{2}.$$ Addendum: Note that the above argument can be easily generalized.


Here is a hint. You may begin by substituing $x=\varphi \tan u$. Then play algebraically at the denominator (by factoring). Don't forget that

$$1+\tan^2 u =\sec^2 u$$

  • $\begingroup$ After $u$-substitution, I get this integral, which gives us the correct answer if we take the integral from $u=0$ to $u=\frac \pi 2$, but I don't see how this integral is easier unless you know the identity of $\frac{a}{a+\tan^2(u)}$. $\endgroup$ – Noble Mushtak Jun 12 '16 at 12:37
  • $\begingroup$ A sorry you were asking for a more easir way. I read an alternative way. My bad :) $\endgroup$ – Tolaso Jun 12 '16 at 12:41
  • $\begingroup$ No problem. I am not the original questioner, I was just commenting here to check if I did what you were hinting at. $\endgroup$ – Noble Mushtak Jun 12 '16 at 12:50

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.