1
$\begingroup$

I am trying to show that density function of standard Cauchy distribution is well defined: that $$\int_\mathbb{R} \frac{1}{\pi(1+x^2)}dx=1.$$

I tried the following calculation: $$\int_{-N}^N \frac{1}{\pi(1+z^2)}dz=\int_{-N}^N\frac{i}{2\pi(z+i)}-\frac{i}{2\pi(z-i)}dz$$$$=\frac{i}{2\pi}\log(z+i)\biggr{|}_{-N}^N-\frac{i}{2\pi}\log(z-i)\biggr{|}_{-N}^N$$$$=\frac{i}{2\pi}(\log(N+i)-\log(-N+i))-\frac{i}{2\pi}(\log(N-i)-\log(-N-i))\to 0$$ as $N\to \infty$. The first equality is due to the fact that $$\frac{1}{z+i}-\frac{1}{z-i}=\frac{-2i}{z+1}.$$ The second equality is due to $$\frac{d\log z}{dz}=\frac{1}{z}$$ on a branch of $\log z$.

I know that I should get $1$ for $$\lim_{N\to\infty}\int_{-N}^N \frac{1}{\pi(1+z^2)}dz,$$ but I just couldn't find the mistake in the above calculation.

Where did I make the mistake? Also, how to calculate this integral if one doesn't do it this way?

Edit: following Oliver's comment, I found that I mis-calculated the limit $$\lim_{N\to\infty} \frac{i}{2\pi}(\log(N+i)-\log(-N+i))-\frac{i}{2\pi}(\log(N-i)-\log(-N-i)).$$ Here I am using the principal branch of logarithm, where $\log$ is given by $\log z=\log r+i\theta$. So, the limit should be exactly 1.

$\endgroup$
3
  • 9
    $\begingroup$ The integrand has a well known antiderivative which you can find in any table. Knowing the antiderivative can give you intuition on what kind of substitution to try. $\endgroup$ May 21 at 23:00
  • 4
    $\begingroup$ You are using incorrect branches of logarithm. In any case, this integral can be done by recalling that $D_x \arctan(x)=\frac{1}{1+x^2}$ $\endgroup$ May 22 at 0:16
  • 1
    $\begingroup$ Thank you Oliver for pointing out the mistake, guess I didn't understand branches of logarithms that well. Also, thanks Ninad. $\endgroup$
    – kid111
    May 22 at 0:56

1 Answer 1

1
$\begingroup$

Since ${d\over dx}(\tan^{-1}{x})={1\over 1+x^2}$ then we have:

$$\int_\mathbb{R}{1\over\pi(1+x^2)}dx=\lim_{N\rightarrow\infty}{1\over \pi}\int_{-N}^{N}{1\over 1+x^2}dx={1\over \pi}\lim_{N\rightarrow\infty}[\tan^{-1}N-\tan^{-1}(-N)]$$ $$={1\over \pi}\lim_{N\rightarrow\infty}[\tan^{-1}N+\tan^{-1}(N)]={2\over \pi}\lim_{N\rightarrow\infty}\tan^{-1}(N)={2\over\pi}\cdot{\pi\over 2}=1$$

$\endgroup$
3
  • 4
    $\begingroup$ A small nitpick. The double sided improper integral is defined when both the limits $\displaystyle\lim_{M \to\infty}\int_{0}^{M}\frac{1}{\pi(1+x^{2})}\,dx$ and $\displaystyle\lim_{N\to\infty}\int_{-N}^{0}\frac{1}{\pi(1+x^{2})}\,dx$ exists finitely and you define $\int_{-\infty}^{\infty}\frac{1}{\pi(1+x^{2})}\,dx$ as their sum. If the integral exists then we can write it as $\displaystyle \lim_{N\to\infty} \int_{-N}^{N}$ $\endgroup$ May 22 at 7:36
  • $\begingroup$ @Mr.GandalfSauron Yes, this is most certainly true. I was debating on including this detail, but decided to go with the p.v, based on what was written in the original post; good looking out! $\endgroup$ May 22 at 8:44
  • $\begingroup$ The integrated function is positive, so $\int\limits_{-N}^N$ is fine. $\endgroup$ May 24 at 14:22

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.