3

Prove that $f(x)=\sum_\limits{n=1}^{\infty} \frac{1}{x^2+n^2}$ is differentiable on $\mathbb{R}$.


Solution:

Let $I=(-a,a)$. Each partial sum $f_k(x)=\sum_1^k \frac{1}{x^2+k^2}$ is differentiable on $I$. Also, $f(0)$ is convergent. Since for $x\in I$

$$|\frac{-2x}{(x^2+k^2)^2}| \le \frac{2a}{(x^2+k^2)^2}$$ and $\sum_1^{\infty}\frac{2a}{(x^2+k^2)^2}$ is convergent, we have that $f'$ exists on $I$ and $f'_k$ is uniformly convergent to $f'$ on $I$. Since $a$ is arbitrarly large, the claim follows. Is this correct?

luka5z
  • 6,359
  • In general, from uniform convergence of differentiable functions doesn't follow neither that the limit function is differentiable nor that, in case the limit actually is differentiable, the series of derivatives converges to the derivative of the limit, and even less uniformly. Under certain conditions, though, the other way around is true. – DonAntonio Feb 28 '16 at 15:09
  • May I ask what $\sum_1^k$ means? – johnny09 Feb 28 '16 at 16:48

1 Answers1

10

Your approach looks fine. The theorem you need to justify your claims is the following (see for example Theorem 2 in this note):

  • Term by term differentiation theorem: If $\sum f_k(x_0)$ converges at some point $x_0\in [a,b]$ and the series of derivatives $g(x) = \sum f_k'(x)$ converges uniformly on $[a,b]$ then the series $f(x) = \sum f_k(x)$ converges uniformly for all $x\in[a,b]$ and $f(x)$ is differentiable with $f'(x) =g(x)$.

Below is two alternative derivations.


From the definition of the derivative we have $$f'(x) = \lim_{h\to 0}\frac{f(x+h) -f(x)}{h} = -\lim_{h\to 0}\sum_{n=1}^\infty\frac{2x+h}{((x+h)^2+n^2)(x^2+n^2)}$$

The summand satisfies $$\left|\frac{2x+h}{((x+h)^2+n^2)(x^2+n^2)}\right| \leq \frac{2|x|+1}{n^4}$$ for $h<1$ so the series above converges uniformly by Weierstrass M-test and we can therefore move the limit $h\to 0$ inside the summation to get

$$f'(x) = -\sum_{n=1}^\infty\lim_{h\to 0}\frac{2x+h}{((x+h)^2+n^2)(x^2+n^2)} = -\sum_{n=1}^\infty \frac{2x}{(x^2+n^2)^2}$$

Since the series above converges for all $x\in\mathbb{R}$ it follows that $f(x)$ is differentiable on $\mathbb{R}$.


Another way to show it is by evaluating the series in closed form

$$f(x) = \left\{\matrix{\dfrac{\pi \coth (\pi x)}{2 x}-\dfrac{1}{2 x^2} & x\not =0\\\dfrac{\pi^2}{6} & x=0}\right.$$

from which it's not hard to see, apart from $x=0$ which requires some work, that the resulting function is differentiable on $\mathbb{R}$.

VIVID
  • 11,604
Winther
  • 24,478
  • 2
    You might consider defining $g(x)=\sum f'_k(x)$. ;-) - Mark – Mark Viola Feb 28 '16 at 16:15
  • 2
    Perfect answer, imo :) – Imago Feb 28 '16 at 16:16
  • @Dr.MV Thanks!. – Winther Feb 28 '16 at 16:17
  • 1
    My pleasure. And +1 for the solution. By the way, one of the issues that I believe that the OP was questioning is a subtle one. We start with a finite and arbitrary interval $(-a,a)$ for which UC of the "derived" series was shown. The conclusion is that the original series is differentiable on $(-\infty,\infty)$ and that its derivative is the "derived" series there. So, even though UC was established for finite intervals, the conclusion applies over the entire set of reals. Subtle, but might be worth a mention. - Mark – Mark Viola Feb 28 '16 at 16:21
  • @Winther Apology. I took the liberty of editing and hadn't seen that you had already edited. – Mark Viola Feb 28 '16 at 16:29