Let $P_n(x) = \frac{n}{1+n^2x^2}$. Prove that for every $n\in\mathbb{N}$ $$\int_{-\infty}^\infty P_n(x) \, dx = \frac{\pi}{n}$$

And for every $\delta > 0$:

$$\lim_{n\to\infty} \int_\delta^\infty P_n(x) \, dx = \lim_{n\to\infty} \int_{-\infty}^{-\delta} P_n(x) = 0$$

This question is Fourier-series-oriented and maybe has something with Fejér's theorem

I'd be glad to get a guidance.


  • 2
    $\begingroup$ For the first part, the form of the integrand suggests the substitution $nx = \tan \theta$. $\endgroup$ – Travis Willse Mar 4 '15 at 9:52
  • $\begingroup$ When $m>n$, $P_m(x)<P_n(x)$ except at $0$. So it's not plausible that the integral is independent of $n$. $\endgroup$ – user99914 Mar 4 '15 at 9:53


$$\int\frac1{1+(nx)^2}dx=\frac1n\int\frac{n\;dx}{1+(nx)^2}=\frac1n\arctan nx+C$$

BTW, the above shows that apparently you forgot a factor $\;\frac1n\;$ in the result of your integral

Addition: for every $\;\delta >0\;$ :

$$\int_\delta^\infty P_n(x) dx=\lim_{b\to\infty}\left.\frac1n\arctan nx\right|_\delta^b=\frac1n\left(\frac\pi2-\arctan\delta\right)\xrightarrow[\delta\to 0]{}\frac\pi{2n}$$

and now you can take the limit when $\;n\to\infty\;$ and get zero indeed.

  • $\begingroup$ Thanks. You're absolutely right. I've corrected it $\endgroup$ – AlonAlon Mar 4 '15 at 9:57
  • $\begingroup$ For the second part:$$\lim_{n\to\infty} \int_\delta^\infty \frac{n}{1+(nx)^2} \, dx = [t= nx] = \lim_{t\to\infty} \int_{n\delta}^\infty \frac{1}{1+t^2} dt = \arctan(\infty)-\lim_{t\to\infty}\arctan(\delta n) \\= \arctan(\infty)-\arctan(\infty) = 0$$ $\endgroup$ – AlonAlon Mar 4 '15 at 10:04
  • $\begingroup$ Do you concur? :) $\endgroup$ – AlonAlon Mar 4 '15 at 10:04
  • $\begingroup$ @Alon, I can't understand the very first expression on the left (and neither the next one). Did you forget to add an integral sign there? $\endgroup$ – Timbuc Mar 4 '15 at 10:05
  • 1
    $\begingroup$ @AlonAlon Check what I added to my answer. $\endgroup$ – Timbuc Mar 4 '15 at 10:19

Recall from calculus that $$ \frac{d}{d\theta} \arctan \theta = \frac{1}{1+\theta^2}. $$

Therefore $$ \frac{d}{d\theta} \left(\frac{1}{n}\arctan n\theta \right)= \frac{1}{1+n^2\theta^2}. $$ The fundamental theorem of calculus then informs us that for any $t_1$ and $t_2$ $$ \int_{t_1}^{t_2}\frac{1}{1+n^2\theta^2} \; d \theta = \frac{1}{n}(\arctan n t_1 - \arctan n t_2). $$ Note that $\lim_{\theta \rightarrow \pm \infty} \arctan \theta = \pm \pi/2$. So upon taking the limits $t_1 \rightarrow - \infty$ and $t_2 \rightarrow \infty$, we find that that $$ \int_{- \infty}^{\infty}\frac{1}{1+n^2\theta^2} = \frac{1}{n}(\pi/2 - (-\pi/2)) = \frac{\pi}{n}. $$

Using the above formula for the integral on any interval $[t_1,t_2]$ we may also readily compute the two latter two integrals. These will be equal to $$ \frac{\pi}{2n} - \frac{1}{n}\delta $$ and $$ - \frac{1}{n}\delta - \frac{\pi}{2n}, $$ respectively. Taking $n \rightarrow \infty$ then gives the desired result.


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.