2
$\begingroup$

I am trying to determine whether the series $\displaystyle\sum_{n=2}^{+\infty} \ln\left(1-\frac{(-1)^n}{\sqrt n}\right)$ converges or not.

I have tried using a Taylor series of the summand, which gives $$v_n = -\frac{(-1)^n}{\sqrt n} - \frac{1}{2n} + O\left(\frac{1}{n\sqrt n}\right)$$ whose series does not converge because the first term gives an alternating series, the big-O is absolutely convergent but the middle term is the harmonic series, and is thus equivalent to $-\frac{1}{2}\ln n$. Thus the series diverges and the partial sum tends to negative infinity.

However, Wolfram Alpha tells me the exponential of the series, the infinite product $\displaystyle\prod_{n=2}^{+\infty}\left(1-\frac{(-1)^n}{\sqrt n}\right)$ has a nonzero limit, which contradicts that.

Have I been doing something wrong ?

$\endgroup$
  • $\begingroup$ Maybe you mistyped the Wolfram Alpha input. The series indeed diverges to $-\infty$. Something you definitely did wrong was to write $\frac{1}{2k}$ instead of $\frac{1}{2n}$ in the expansion of $v_n$. $\endgroup$ – Daniel Fischer Oct 4 '15 at 20:17
  • $\begingroup$ Well, the product of two consecutive terms of the form $1-\frac{(-1)^n}{\sqrt{n}}$ is $1-\frac{1}{n}+O\left(\frac{1}{n\sqrt{n}}\right)$, and your proof of divergence looks perfectly fine. $\endgroup$ – Jack D'Aurizio Oct 4 '15 at 20:22
  • $\begingroup$ @DanielFischer: Wolfram|Alpha does give a non-zero limit. $\endgroup$ – joriki Oct 4 '15 at 20:42
  • $\begingroup$ @DanielFischer Whoops. I copied straight over from a sheet of paper where it was $v_k$. Thank you I edited the original post. $\endgroup$ – Groovy. Oct 4 '15 at 21:02
  • $\begingroup$ Good, so @joriki's link proves that Wolfram Alpha gives a wrong answer. While that doesn't positively prove that you didn't mistype, we can relegate that to a merely theoretical possibility and conclude that you haven't done anything wrong here. The error is on the other side of the internet connection. $\endgroup$ – Daniel Fischer Oct 4 '15 at 21:07
1
$\begingroup$

Wolfram Alpha was wrong, and my proof was correct. We indeed have $$ \prod_{n=2}^\infty \left(1-\frac{(-1)^n}{\sqrt n}\right) = 0 $$ as $\displaystyle\sum _{k=2}^n \ln\left(1-\frac{(-1)^k}{\sqrt k}\right) = -\frac{1}{2}\ln(n) + \mathrm O(1)$ as $n $ approaches infinity.

$\endgroup$
  • $\begingroup$ interesting path that could have been taken is to take: $$\exp\sum_{k=2}^n \ln\left(1-\frac{(-1)^k}{\sqrt k}\right)) = \prod_{n=2}^{\infty}e^{ln(1-\frac{(-1)^{n}}{\sqrt(n)}}$$ Now all one, has to show: $$\lim_{N \rightarrow \infty}\prod_{n=2}^{N}e^{ln(1-\frac{(-1)^{n}}{\sqrt(n)}} = 0$$ $\endgroup$ – Zophikel Dec 22 '17 at 23:45
  • $\begingroup$ I think this can be done via a Stirling's approximation $\endgroup$ – Zophikel Dec 22 '17 at 23:45

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.