4
$\begingroup$

I am attempting to prove if the below expression converges or diverges:

$$\sum_{n=2}^{\infty} \frac{1}{n (\ln n)^2}$$

I decided to try using the limit convergence test. So then, I set $f(n) = \frac{1}{n (\ln n)^2}$ and $g(n) = \frac{1}{n}$ and did the following:

$$ \lim_{n \to \infty} \frac{f'(n)}{g'(n)} = \lim_{n \to \infty} \frac{-\frac{\ln(n) + 2}{n^2 (\ln n)^3}}{-\frac{1}{n^2}} = \lim_{n \to \infty} \frac{\ln(n) + 2}{n^2 (\ln n)^3} \cdot n^2 = \lim_{n \to \infty} \frac{1}{(\ln n)^2} + \frac{2}{(\ln n)^3} = 0 $$

I know that $\sum \frac{1}{n}$ diverges due to properties of harmonic series, and so concluded that my first expression $\frac{1}{n (\ln n)^2}$ also must diverge.

However, according to Wolfram Alpha, and as suggested by this math.SE question, the expression actually converges.

What did I do wrong?

$\endgroup$
9
  • 1
    $\begingroup$ You can use Cauchy's condensation test. $\endgroup$ – Amihai Zivan Nov 18 '12 at 21:50
  • $\begingroup$ Your $f(x),g(x)$ do not depend on $x$, hence $f'(x)=g'(x)=0$. But you derive them $dn$, where $n$ is discrete. You should take much more care when doing things like that. $\endgroup$ – Dennis Gulko Nov 18 '12 at 21:52
  • 2
    $\begingroup$ The integral test works well too. $\endgroup$ – mrf Nov 18 '12 at 21:55
  • $\begingroup$ is it $\log(x^2)$ or $(\log x)^2$? $\endgroup$ – user31280 Nov 18 '12 at 22:04
  • $\begingroup$ @F'OlaYinka -- the latter $\endgroup$ – Michael0x2a Nov 18 '12 at 22:10
9
$\begingroup$

In the limit comparison test, $\sum_n f(n)$ and $\sum_n g(n)$ both will behave the same iff $$\lim_{n \to \infty} \dfrac{f(n)}{g(n)} = c \in (0,\infty)$$ In your case, the limit is $0$ and hence you cannot conclude anything.

$\endgroup$
6
$\begingroup$

You can use integral test for convergence. Since the integral $$\int_2^\infty\dfrac{1}{x\log^2 x}dx$$ converges, the series $\displaystyle{\sum_{n=2}^{\infty} \frac{1}{n (\ln n)^2}}$ converges.

$\endgroup$
3
  • 1
    $\begingroup$ Thank you -- this is what I ended up using, but I decided to accept Marvis's answer because it helped show me why the limit test wasn't working here. $\endgroup$ – Michael0x2a Nov 18 '12 at 22:12
  • $\begingroup$ The integral converges to 2, but is it correct to say that the series also converges to 2, or can we not make that conclusion? $\endgroup$ – stackdsewew Jun 12 '16 at 9:09
  • $\begingroup$ @anna_xox I found that it converges to $\frac{1}{ln(2)}$ - did I calculate wrong? $\endgroup$ – Elad Avron Jun 27 '16 at 10:17
5
$\begingroup$

Expanding on the accepted answer, you should use the Cauchy condensation test: For a positive non-increasing sequence $f(n)$, the series $\sum_{n=0}^\infty f(n)$ converges if and only if the series $\sum_{n=0}^\infty2^nf(2^n)$ does.
In your case, $f(n)=\frac{1}{n\ln^2n}$, so $$2^nf(2^n)=2^n\frac{1}{2^n\ln^22^n}=\frac{1}{n^2\ln^22}=\frac{1}{\ln^22}\frac{1}{n^2}$$ which clearly converges.

$\endgroup$

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.