9
$\begingroup$

$$\int_{0}^{\infty}\frac{4x^2}{(x^4+2x^2+2)^2}dx=\frac{\pi}{15}$$

$$\int_{0}^{\infty}\frac{4x^2}{[(1+(1+x^2)^2]^2}dx=\frac{\pi}{15}$$

$u=\tan(z)$ $\rightarrow$ $du=\sec^2(z)$

$u$ $\rightarrow \infty$, $\tan(z)=\frac{\pi}{2}$

$u$ $\rightarrow 0$, $\tan(z)=0$

$$\int_{0}^{\pi \over 2}\frac{4\tan^2(z)}{[(1+(1+\tan^2(z))^2]^2}\frac{du}{\sec^2(z)}=\frac{\pi}{15}$$

$1+\tan^2(z)=\sec^2(z)$

$$\int_{0}^{\pi \over 2}\frac{4\sin^2(z)}{[(1+\sec^4(z)]^2}dx=\frac{\pi}{15}$$

$$\int_{0}^{\pi \over 2}\frac{4\sin^2(z)}{[(1+i\sec^2(z))(1-i\sec^2(z))]^2}dx=\frac{\pi}{15}$$

Hopeless! Any suggestion?

Try again $$\int_{0}^{\pi \over 2}\frac{4\sin^2(z)}{[(1+\sec^4(z)]^2}dx=\frac{\pi}{15}$$

$$\int_{0}^{\infty}\frac{\sin^2(2z)\cos^6(z)}{(1+\cos^4(z))^2}dx=\frac{\pi}{15}$$

$\sin^2(2z)=\frac{1-\cos(4z)}{2}$

No more I gave up. Any hints?

$\endgroup$
11
  • 1
    $\begingroup$ Are you familiar with the residue theorem? $\endgroup$
    – joriki
    May 30, 2016 at 22:29
  • $\begingroup$ Alternatively, you could factor the denominator and do a partial fraction expansion. $\endgroup$
    – M10687
    May 30, 2016 at 22:30
  • $\begingroup$ @joriki no, I have no idea about residue theorem $\endgroup$ May 30, 2016 at 22:36
  • $\begingroup$ Are you interested in a solution using the residue theorem? $\endgroup$
    – joriki
    May 30, 2016 at 22:37
  • 1
    $\begingroup$ That is close to $\frac{\pi}{15}$, but the value of the integral is not $\frac{\pi}{15}$, so I suggest to modify the question title. $\endgroup$ May 30, 2016 at 23:28

2 Answers 2

6
$\begingroup$

Hint. One may write $$ \begin{align} \int_{0}^{\infty}\frac{4x^2}{(x^4+2x^2+2)^2}dx&=4\int_0^{\infty}\frac1{\left(\left(x-\frac{\sqrt{2}}x\right)^2+2+2\sqrt{2}\right)^2}\:\frac{dx}{x^2} \\\\&=2\sqrt{2}\int_0^{\infty}\frac1{\left(\left(x-\frac{\sqrt{2}}x\right)^2+2+2\sqrt{2}\right)^2}\:dx \quad (x \to \sqrt{2}/x) \\\\&=\frac1{\sqrt{2}}\int_{-\infty}^{\infty}\frac{\left(1+\frac{\sqrt{2}}{x^2}\right)}{\left(\left(x-\frac{\sqrt{2}}x\right)^2+2+2\sqrt{2}\right)^2}\:dx \\\\&=\frac1{\sqrt{2}}\int_{-\infty}^{\infty}\frac{d\left(x-\frac{\sqrt{2}}x\right)}{\left(\left(x-\frac{\sqrt{2}}x\right)^2+2+2\sqrt{2}\right)^2} \\\\&=\frac1{\sqrt{2}}\int_{-\infty}^{\infty}\frac{du}{(u^2+2+2\sqrt{2})^2} \\\\&=\color{red}{\frac14 \sqrt{5\sqrt{2}-7} \:\pi} \\\\ \end{align} $$

where we have made $u:=\sqrt{2+2\sqrt{2}}\:\sinh v$ to get the last step.

$\endgroup$
3
  • $\begingroup$ (+1) I agree with your computation, $\frac{\pi}{15}$ is not the actual value of the integral. $\endgroup$ May 30, 2016 at 23:27
  • $\begingroup$ @JackD'Aurizio Thank you. $\endgroup$ May 31, 2016 at 0:15
  • $\begingroup$ First of all @OlivierOloa thank for the correct answer. I admired your first step of transformation, which I understand clearly as to what follow i am quite lose. If you can provide me more detail. Thank you. $\endgroup$ May 31, 2016 at 4:03
3
$\begingroup$

Here is an approach using contour integration in case anyone is interested. An effort has been made to use pen-and-paper type manipulations only. These are simple yet demand a certain care with the algebra. Suppose we seek to verify that

$$\int_0^\infty \frac{4x^2}{(x^4+2x^2+2)^2} dx = \frac{\pi}{4} \sqrt{5\sqrt{2}-7}$$

or alternatively

$$\int_0^\infty \frac{x^2}{(x^4+2x^2+2)^2} dx = \frac{\pi}{16} \sqrt{5\sqrt{2}-7}.$$

We use a semicircular contour in the upper half plane with two straight components $\Gamma_0$ and $\Gamma_1$ on the positive and negative real axis and having radius $R$ ($\Gamma_2.$)

The denominator here is $$((x^2+1)^2+1)^2$$ so the poles are double and located at

$$\rho_{0,1,2,3} = \pm\sqrt{-1\pm i}.$$

We convert this to polar form in order to determine which poles are in the upper half plane, getting

$$\pm\sqrt{\sqrt{2} \exp(\pi i \pm \pi i/4)} = \sqrt[4]{2} \exp(\pi i/2 \pm \pi i/8 + \pi i/2 \pm \pi i/2) \\ = \sqrt[4]{2} \exp(\pi i \pm \pi i/8 \pm \pi i/2).$$

Fortunately we can see by inspection that only the two poles $$\rho_0 = \sqrt[4]{2} \exp(\pi i - \pi i/8 - \pi i/2) = \sqrt[4]{2} \exp(3\pi i/8) \\ \quad\text{and}\quad \rho_1 = \sqrt[4]{2} \exp(\pi i + \pi i/8 - \pi i/2) = \sqrt[4]{2} \exp(5\pi i/8)$$

are inside the contour (arguments are $3\pi/8$ and $5\pi/8$, the other two are at $-3\pi /8$ and $-5\pi /8.$)

For the residue we get

$$\frac{1}{2\pi i} \int_{|z-\rho_0|=\epsilon} \frac{z^2}{(z^4+2z^2+2)^2} \; dz.$$

In order to get a pole that is amenable to easy algebra we introduce $w = z\exp(-3\pi i/8)/\sqrt[4]{2}$ and $z = w\exp(3\pi i/8)\sqrt[4]{2}$ which maps $\rho_0$ to $1$ so we obtain

$$\exp(3\pi i/4+3\pi i/8)\sqrt{2}\sqrt[4]{2} \\ \times \frac{1}{2\pi i} \int_{|w\exp(3\pi i/8)\sqrt[4]{2}-1|=\epsilon} \frac{w^2}{(-2iw^4+2w^2(-1+i)+2)^2} \; dw \\ = - \exp(9\pi i/8) \frac{2^{3/4}}{4} \frac{1}{2\pi i} \int_{|w-1|=\gamma} \frac{w^2}{(w-1)^2\times (w^2-i)^2(w+1)^2} \; dw.$$

The residue is thus given by

$$- \exp(9\pi i/8) \frac{2^{3/4}}{4} \lim_{w\rightarrow 1} \left(\frac{w^2}{(w^2-i)^2(w+1)^2}\right)' \\ = - \exp(9\pi i/8) \frac{2^{3/4}}{4} \lim_{w\rightarrow 1} \left(\frac{2w}{(w^2-i)^2(w+1)^2} \\ - \frac{w^2}{(w^2-i)^4(w+1)^4} (2(w^2-i) 2w (w+1)^2 + (w^2-i)^2 2(w+1))\right).$$

This works out to $$-\exp(9\pi i/8) \frac{2^{3/4}}{4} \times \frac{1}{8} (2-i) = i\exp(-3\pi i/8) \frac{2^{3/4}}{4} \times \frac{1}{8} \sqrt{5} \exp(-i\beta)$$

where $2-i = \sqrt{5}\exp(-i\beta).$

Continuing with the second pole we we introduce $w = z\exp(-5\pi i/8)/\sqrt[4]{2}$ and $z = w\exp(5\pi i/8)\sqrt[4]{2}$

and obtain

$$\exp(5\pi i/4+5\pi i/8)\sqrt{2}\sqrt[4]{2} \\ \times \frac{1}{2\pi i} \int_{|w\exp(5\pi i/8)\sqrt[4]{2}-1|=\epsilon} \frac{w^2}{(2iw^4+2w^2(-1-i)+2)^2} \; dw \\ = - \exp(15\pi i/8) \frac{2^{3/4}}{4} \frac{1}{2\pi i} \int_{|w-1|=\gamma} \frac{w^2}{(w-1)^2\times (w^2+i)^2(w+1)^2} \; dw.$$

This is the same as in the previous pole except the sign in the $w^2-i$ term has been flipped. Re-using the derivative thus yields

$$-\exp(15\pi i/8) \frac{2^{3/4}}{4} \times \frac{1}{8} (2+i) = i\exp(3\pi i/8) \frac{2^{3/4}}{4} \times \frac{1}{8} \sqrt{5} \exp(i\beta).$$

Adding the two residues we thus obtain

$$\frac{2^{3/4}}{32}\times\sqrt{5}\times 2i\cos(\beta+3\pi /8).$$

Returning to the main computation, on the part of the contour that is on the negative real axis which is $\Gamma_1$ we trivially obtain

$$\int_{-R}^0 \frac{x^2}{(x^4+2x^2+2)^2} \; dx$$

which yields with $z=-x$

$$- \int_R^0 \frac{z^2}{(z^4+2z^2+2)^2} \; dz = \int_{\Gamma_0} \frac{z^2}{(z^4+2z^2+2)^2} \; dz.$$

Finally we have by the ML bound for the semicircular component

$$\lim_{R\rightarrow\infty} \left|\int_{\Gamma_2} \frac{z^2}{(z^4+2z^2+2)^2} \; dz\right| \le \lim_{R\rightarrow\infty} 2\pi R/2 \times \frac{R^2}{(R^4-2R^2+2)^2} = 0.$$

It follows that

$$\int_0^\infty \frac{x^2}{(x^4+2x^2+2)^2} \; dx = \frac{1}{2}\times 2\pi i \times \frac{2^{3/4}}{32}\times\sqrt{5}\times 2i\cos(\beta+3\pi /8) \\ = -\frac{\pi}{16} 2^{3/4} \sqrt{5} \cos(\beta+3\pi /8).$$

To manipulate this to match the form in the introduction we use angle sum and half-angle formulae as in

$$\sqrt{5}\cos(\beta+3\pi /8) = \sqrt{5}\cos\beta\cos(3\pi /8) - \sqrt{5}\sin\beta\sin(3\pi /8) \\ = 2\cos(3\pi /8) - \sin(3\pi /8).$$

As we are integrating a function that is never negative on the integration interval we see that the sign on this last term must be negative. Observe that

$$\cos(3\pi/8) = \sqrt{\frac{1+\cos(3\pi/4)}{2}} = \sqrt{\frac{1-\sqrt{2}/2}{2}}$$

and $$\sin(3\pi/8) = \sqrt{\frac{1-\cos(3\pi/4)}{2}} = \sqrt{\frac{1+\sqrt{2}/2}{2}}.$$

Squaring we obtain

$$4 \frac{1-\sqrt{2}/2}{2} + \frac{1+\sqrt{2}/2}{2} - 4 \sqrt{\frac{1-2/4}{4}} = \frac{5}{2} - \frac{3}{4}\sqrt{2}-\sqrt{2} \\ = \frac{5}{2} - \frac{7}{4}\sqrt{2}.$$

We thus have for the end result

$$-\frac{\pi}{16} 2^{3/4} \times - \sqrt{\frac{5}{2} - \frac{7}{4}\sqrt{2}} = \frac{\pi}{16} \times \sqrt{\frac{5}{2}2^{3/2} - \frac{7}{4}2^2} \\ = \frac{\pi}{16} \times \sqrt{5\sqrt{2}-7}.$$

This is the claim.

$\endgroup$
2
  • $\begingroup$ (+1)Marko Riedel. This is passed my pay grade. Is this a bachelor or a master level of maths? $\endgroup$ Jun 1, 2016 at 4:03
  • $\begingroup$ Thank you for the upvote, my solution using complex variables is admittedly somewhat more involved than the other approaches. I would say it is about the level of a junior / senior introductory course in complex variables. $\endgroup$ Jun 1, 2016 at 10:52

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .