3

I am working through this problem:- Show that $\int_0^ \infty \frac{1}{1+x^n} dx= \frac{ \pi /n}{\sin(\pi /n)}$ , where $n$ is a positive integer.

I follow it all, except for part (3) - I think this must a technique I haven't encountered yet, why does the integral on the incoming ray tend to $ -e^{\frac{2 \pi i}{n}} \int_{0}^{\infty}\frac{dx}{1+x^n} $ ?

I tried - because the function is analytic, it is path independent - $ \lim_{{R}\to{\infty}} \int_{R}^{0}f(z)dz = f(0) - f(R) $. Not promising.

What I can sort of see, might be to paramerterize $ z= Re^{\frac{2 \pi i}{n}}, dz=e^{\frac{2 \pi i}{n}}dR $ and then find $ \lim_{{R}\to{\infty}} e^{\frac{2 \pi i}{n}} \int_{0}^{R}\frac{dR}{1+R^ne^{2 \pi i}}$ but thats no better than the original problem. A hint on the approach would be very helpful, thanks

Ognik
  • 43
  • Thanks @Dr. MV, just out of interest, I used a 3 circle technique I came across elsewhere for $C_2$ arc, to illustrate that inequality. However - and I'm sure I'm missing the blindingly obvious here - could you explain the parameterization for $C_3$ please? I don't get it unfortunately. – Ognik Aug 14 '15 at 10:39
  • Actually what confused me was that your parameterisation for $C_2$ was effectively the same as mine, except I used R instead of t - so I have been looking for a difference where there wasn't one. The 'trick' I have finally seen, lies in not having to do the integration - thanks again for the help.

    PS: If using comments like this is not the right way for me to reply, please let me know, I am relatively new to stack exchange maths.

    – Ognik Aug 14 '15 at 21:54

1 Answers1

3

It might be instructive just to walk through the entire process. So, to that end here we go.

First, we note on the closed contour $C$, we can parameterize the three connected contours $C_1$, $C_2$, and $C_3$ as follows:

On $C_1$, we are on the real axis and have $z=t$ and $dz=dt$, where $t$ starts at $0$ and ends at $R$.

On $C_2$, we are on a circular arc with $z=Re^{it}$ and $dz=iRe^{it}dt$, where $t$ starts at $0$ and ends at $2\pi/n$.

On $C_3$, we are on a straight line segment with $z=e^{i2\pi/n}\,t$ and $dz=e^{i2\pi/n}\,dt$, where $t$ starts at $R$ and ends at $0$.


Now that we have our parameterizations, let's proceed to evaluate the integral contributions to each. First, as $R\to \infty$ the contribution to the closed contour integral from the integration over the circular arc $C_2$ vanishes. To see this, we write

$$\begin{align} \left|\int_0^{2\pi/n}\frac{1}{1+R^ne^{int}}iRe^{it}dt\right|&\le\int_0^{2\pi/n}\frac{1}{|1+R^ne^{int}|}Rdt\\\\ &\le\int_0^{2\pi/n}\frac{1}{R^n-1}Rdt\\\\ &= \frac{2\pi}{n}\frac{R}{R^n-1}\\\\ &\to 0 \end{align}$$

as $R\to \infty$, for integer $n>1$.


Having evaluated the integral over $C_2$ as zero, we must have that the sum of the contributions of integrations over $C_1$ and $C_3$ must equal $2\pi i$ times the residue of the function $\frac{1}{1+z^n}$ at the single enclosed pole at $z=e^{i\pi/n}$. Therefore, we have that

$$\begin{align} \int_0^{\infty}\frac{1}{1+t^n}dt+\int_{\infty}^{0}\frac{1}{1+t^n}e^{i2\pi/n}dt&=(1-e^{i2\pi/n})\int_0^{\infty}\frac{1}{1+x^n}\,dx\\\\ &=2\pi i \text{Res}\left(\frac{1}{1+z^n},z=e^{i\pi/n}\right) \end{align}$$

The residue from the single enclosed pole is given by

$$\begin{align} \text{Res}\left(\frac{1}{1+z^n},z=e^{i\pi/n}\right)&=\lim_{z\to e^{i\pi/n}}\frac{(z-e^{i\pi/n})}{1+z^n}\\\\ &=\lim_{z\to e^{i\pi/n}}\frac{1}{nz^{n-1}}\\\\ &=\frac{1}{-ne^{-i\pi/n}} \end{align}$$


Finally, putting it all together, we find that

$$\int_0^{\infty}\frac{1}{1+x^n}dx=\frac{-2\pi i}{ne^{-i\pi/n}}\frac{1}{1-e^{i2\pi/n}}=\frac{\pi/n}{\sin(\pi/n)}$$

as expected!

Mark Viola
  • 179,405
  • 1
    One small edit might be needed on the line before "Putting it all together.." namely the dashed line fraction – Leucippus Aug 14 '15 at 04:11
  • @Leucippus Thanks for the catch!! +1 – Mark Viola Aug 14 '15 at 04:11
  • As always Great Solution! – Leucippus Aug 14 '15 at 04:15
  • @Leucippus As always, thank you!! And I've noticed you've been busy posting some impressive answers to very tough integrals. – Mark Viola Aug 14 '15 at 04:17
  • occasionally my brain kicks in, or...the bag of math tricks is filled with good ones. – Leucippus Aug 14 '15 at 04:21
  • @Leucippus On "tricks," ... A friend of mine who's written many texts on math & physics, spoke to a coauthor, a Russian mathematician, about some extremely difficult problems (in inequalities since my friend wrote a book on applied inequalities) he's seen young students solve quickly. The coauthor told him that some of these students were coached from 10-11 years old on "tricks" for the Olympia. 1 trick per day, 10 problems per trick. So with about 2,000 tricks and a good memory they can solve some ridiculously hard problems. Deviate from a trick-oriented problem and they often crumble. – Mark Viola Aug 14 '15 at 04:33
  • "tricks" are good to know, but if not applied everywhere then they become they can not become "tools of the trade". – Leucippus Aug 14 '15 at 05:31
  • 1
    @Leucippus All of us have learned tricks. The key is to understand principles to be able to adapt when the tricks run out. – Mark Viola Aug 14 '15 at 05:39