How we could know that for the function $f(x)$ below if there exist (not exist) a constant $C>0$ such that $f(x)\geq C$ for all $x\in \mathbb R$, $$f(x)=\sum_{n=1}^{\infty}\frac{1}{(x-n)^{2}+1}$$ (i.e., $\lim_{x\to\pm\infty}f(x)\neq 0$)

EDIT: I got something which I'm not sure if its correct or not:

If $x>0$ is too large then there is $n_{o}$ such that $x<n_{o}$, so we split the summation $$\sum_{n=1}^{n_{o}-1}\frac{1}{(x-n)^{2}+1}+\sum_{n=n_{o}}^{\infty}\frac{1}{(x-n)^{2}+1}$$

and for any $n\geq n_{o}$ we have $(x-(n+k))^{2}< (\epsilon+k)^{2}$ in this case the second summation will be finite, where $\epsilon = n_{o}-x$.

  • $\begingroup$ Do you expect the same behavior as $x\to \infty$ and $x\to -\infty$? How might they differ? $\endgroup$ – Jonas Meyer Sep 28 '12 at 0:43
  • $\begingroup$ @Jonas Meyer: I just want to check if both limits are nonzero. $\endgroup$ – Cherly Sep 28 '12 at 0:45
  • $\begingroup$ Cherly: I think I understand that. Interesting. Would you mind sharing where the problem came from? $\endgroup$ – Jonas Meyer Sep 28 '12 at 0:49
  • $\begingroup$ It came up during the class, discussion about interchanging the order of limit and summation. $\endgroup$ – Cherly Sep 28 '12 at 0:59
  • $\begingroup$ Note that the assertion in your last two lines of the question is false: the term on the left goes to infinity when $n$ does, while the right hand side is fixed. $\endgroup$ – Martin Argerami Sep 28 '12 at 1:33

When $x\geq \frac{1}{2}$, there exists a positive integer $k$ such that $|x-k|\leq \frac{1}{2}$. Thus $(x-k)^2\leq \frac{1}{4}$. What does this imply about $f(x)$ for $x>0$, and how does it answer the $x\to \infty$ part?

When $x<0$, $x$ will no longer be so close to positive integers. Note that $f$ is increasing on $(-\infty,1]$, so it suffices to find the sequence limit $$\lim\limits_{k\to\infty}f(-k)=\lim\limits_{k\to\infty}\sum\limits_{n=1}^\infty\dfrac{1}{(k+n)^2+1}.$$ The series $\sum\limits_{n=1}^\infty\dfrac{1}{n^2+1}$ converges, which means that the sequence of partial sums converges, and

$$\sum\limits_{n=1}^\infty\dfrac{1}{(k+n)^2+1}\ \ =\sum\limits_{n=1}^\infty\dfrac{1}{n^2+1}-\sum\limits_{n=1}^{k}\dfrac{1}{n^2+1}\to 0.$$

That is, $f(-k)\to 0$ as $k\to \infty$ because the tail of a convergent series goes to $0$.

  • $\begingroup$ Nice! What about the other limit at $+\infty$? $\endgroup$ – Cherly Sep 28 '12 at 1:36
  • $\begingroup$ @Cherly: I am inviting you to investigate how the $+\infty$ case is handled by the observation in the first paragraph of my answer. $\endgroup$ – Jonas Meyer Sep 28 '12 at 1:37

Related problem: (I). Here is a closed form for the series, in terms of the psi function that may help you analyze your problem

$$ \sum_{n=1}^{\infty}\frac{1}{(x-n)^{2}+1}=-\frac{1}{2}\,i \left( -\psi \left( 1-i-x \right) +\psi \left( 1+i-x \right) \right)\,.$$

Another interesting closed form is when you sum the series from $-\infty$ to $\infty$, you get the following function

$$ \sum_{n=-\infty}^{\infty}\frac{1}{(x-n)^{2}+1}={\frac {\pi \,\cosh \left( \pi \right) \sinh \left( \pi \right) }{ \left( \cosh \left( \pi \right) \right) ^{2}- \left( \cos \left( \pi \,x \right) \right) ^{2}}} $$

  • $\begingroup$ So for the second form the limit at $\pm\infty$ is zero, this is also interesting, thanks! How did you get the $\cosh$ and $\sinh$ functions to be involved here? $\endgroup$ – Cherly Sep 28 '12 at 2:53
  • $\begingroup$ @Cherly: The second function is periodic, and has no limits at $\pm \infty$. $\endgroup$ – Jonas Meyer Sep 28 '12 at 5:09
  • $\begingroup$ Actually, the second function can be used to help answer the question of what happens to the original function at $+\infty$. Note that $f(x)+f(-x) +\frac{1}{x^2+1}$ is periodic and nonconstant, hence has no limits at $\pm\infty$, hence $f$ cannot have limits at both $+\infty$ and $-\infty$. (What I have said here doesn't quite show that $f$ has a positive lower bound in one direction, however, for which the method hinted in my answer, or perhaps use of the Psi functions, may be of more use.) $\endgroup$ – Jonas Meyer Sep 28 '12 at 5:50

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.