Quantcast
Channel: Active questions tagged real-analysis - Mathematics Stack Exchange
Viewing all articles
Browse latest Browse all 9285

A difficult example to show that limits and integrals can't always be switched

$
0
0

I wanted an interesting one, so I started with

$\lim\limits_{N\to\infty}\int\limits_{-\infty}^{\infty} \operatorname{sech}\left(x-\sum\limits_{n=1}^{N}\frac{1}{x+n^{2}}\right)dx$

The idea was that, by Glasser's master theorem, this direction should be easy to evaluate, as it should just become

$\lim\limits_{N\to\infty}\int\limits_{-\infty}^{\infty} \operatorname{sech}(x)dx=\lim\limits_{N\to\infty}\pi=\pi$

Then, the other direction will be

$\int\limits_{-\infty}^{\infty}\lim\limits_{N\to\infty}\operatorname{sech}\left(x-\sum\limits_{n=1}^{N}\frac{1}{x+n^{2}}\right)dx=\int\limits_{-\infty}^{\infty}\operatorname{sech}\left(x-\sum\limits_{n=1}^{\infty}\frac{1}{x+n^{2}}\right)dx$

Using the result from complex analysis that states

$\sum\limits_{k=1}^{\infty}\frac{1}{k^{2}+a^{2}}=-\frac{1}{2a^2}+\frac{\pi}{2a}\coth(\pi a)$

the integral should transform into

$\int\limits_{-\infty}^{0}\operatorname{sech}(x+\frac{1}{2x}+\frac{\pi}{2\sqrt{-x}}\cot(\pi\sqrt{-x}))dx+\int\limits_{0}^{\infty}\operatorname{sech}(x+\frac{1}{2x}-\frac{\pi}{2\sqrt{x}}\coth(\pi\sqrt{x}))dx$

A substitution $-x\mapsto x$ on the first makes it

$\int\limits_{0}^{\infty} \operatorname{sech}\left(x + \frac{1}{2x} - \frac{\pi}{2\sqrt{x}} \cot{\left(\pi\sqrt{x}\right)}\right) + \operatorname{sech}\left(x + \frac{1}{2x} - \frac{\pi}{2\sqrt{x}} \coth{\left(\pi\sqrt{x}\right)}\right)dx$

Now I need to either evaluate it exactly, or show it's not equal to $\pi$. I initially thought Feynman's technique might help. If I consider

$I(p)=\int\limits_{0}^{\infty} \operatorname{sech}\left(x + \frac{1}{2x} - \frac{\pi}{2\sqrt{px}} \cot{\left(\pi\sqrt{px}\right)}\right) + \operatorname{sech}\left(x + \frac{1}{2x} - \frac{\pi}{2\sqrt{px}} \coth{\left(\pi\sqrt{px}\right)}\right)dx$

then I want $I(1)$. The idea was, the terms with $p$ should both vanish almost everywhere as $p\to\infty$, so I would end up with

$\lim\limits_{p\to\infty} I(p)=\int\limits_{0}^{\infty}2\operatorname{sech}(x+\frac{1}{2x})dx$

I would then apply Glasser's master theorem again to get $\lim\limits_{p\to\infty}I(p)=\pi$, and use sign analysis on $I'(p)$ to show it wouldn't take this value for a finite $p$. The issue is that Glasser's master theorem would only apply if I was subtracting the $\frac{1}{2x}$, not adding it, so $\lim\limits_{p\to\infty}I(p)$ may be different from $\pi$. So, I don't know what technique would allow this to progress further, and how we'd find the closed form, if there somehow is one.


Viewing all articles
Browse latest Browse all 9285

Trending Articles



<script src="https://jsc.adskeeper.com/r/s/rssing.com.1596347.js" async> </script>