Does the following series $$ g(x) = \sum_{n=1}^{\infty} \frac{1}{(nx)^2+1} $$ converge uniformly

  • on $x \in (0,1)$

  • on $x\in (1, \infty)$

  • Calculate limit $\lim_{x \rightarrow \infty} g(x)$.

My attempts:

when $x \in (1, \infty)$ we have

$$ \sum_{n=1}^{\infty} \frac{1}{(nx)^2+1} \leq \sum_{n=1}^{\infty} \frac{1}{n^2x^2} \leq \sum_{n=1}^{\infty} \frac{1}{n^2} $$ so series converge uniformly based on Weierstrass Theorem.

when $x \in (0, 1)$ we have

$$ \lim_{n \rightarrow \infty} \frac{1}{(nx)^2+1} = 0 = f(x) \\ f_n \rightarrow f \\ \text{We want to ensure that for every }\epsilon\sup_{x \in (0,1)} \left|\frac{1}{(nx)^2+1} - 0 \right | < \epsilon $$ Lets find its extreme points by differentiating: $$ f_n'(x) = \frac{-2n^2x}{((nx)^2 +1)^2} $$ so the function is strictly decreasing and has may have an extremum when x = 0. But since $x = 0 \notin (0,1)$ we use limit to find values near zero: $$ \lim_{\epsilon \rightarrow 0^{+}} \frac{1}{(n\epsilon)^2+1} = 1 \gt \epsilon $$

So series converge uniformly only when $x > 1$.

Is that a correct reasoning? How can I find its limit? Does it equal the function $f$ it converges to so it's just 0?


I would do it as follows: if your series was uniformly convergent on $(0,1)$, then the sequence $\bigl(f_n(x)\bigr)_{n\in\mathbb N}$, where $f_n(x)=\frac1{(nx)^2+1}$, would converge uniformly to the null function. But it doesn't, since$$(\forall n\in\mathbb N):f_n\left(\frac1n\right)=\frac12.$$

On the other hand, if $N\in\mathbb N$, then$$g(N)=\sum_{n=1}^\infty\frac1{(Nn)^2+1}<\sum_{n=1}^\infty\frac1{(Nn)^2}<\sum_{n=N}^\infty\frac1{n^2}\to_{N\to\infty}0.$$This, together with the fact that $g$ is decreasing on $(1,\infty)$ (since its the sum of decreasing functions), shows that $\lim_{x\to\infty}g(x)=0.$

| cite | improve this answer | |

If you like an overkill, you may notice that $\frac{1}{n^2+a^2}=\int_{0}^{+\infty}\frac{\sin(nt)}{n}e^{-at}\,dt$ for any $a,n\in\mathbb{N}^+$.
It follows that $$ \sum_{n\geq 1}\frac{1}{n^2 x^2+1} = \frac{1}{x^2}\sum_{n\geq 1}\int_{0}^{+\infty}\frac{\sin(nt)}{t} e^{-t/x}\,dt =\frac{1}{x^2}\int_{0}^{+\infty}W(t)e^{-t/x}\,dt$$ where $W(t)$ is the $2\pi$-periodic extension of the function (sawtooth wave) which equals $\frac{\pi-t}{2}$ on $(0,2\pi)$. By explicit integration, it follows that $$ \sum_{n\geq 1}\frac{1}{n^2 x^2+1} = \frac{\pi\coth\frac{\pi}{x}-x}{2x} $$ so the answer to the third point is simply zero. The RHS is unbounded in a right neighbourhood of the origin, hence we cannot have uniform convergence on $(0,1)$. If $x\in(0,1)$ the RHS is actually extremely close to $\frac{\pi-x}{2x}$. On the other hand the uniform convergence over $[1,+\infty)$ is almost trivial since the LHS is a positive, continuous and decreasing function of the $x$ variable.

| cite | improve this answer | |

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.