5
$\begingroup$

I have a sequence of complex numbers $a_1,a_2,...$ such that $a_i \neq -1$. I then have the infinite product $\prod\limits_{n=1}^{\infty}(1+a_n)$ which I know converges to a non zero complex number. I was wondering if this is enough to guarantee the convergence of $\sum\limits_{n=1}^\infty\frac{a_n}{1+a_n}$.

I couldn't come up with an obvious counterexample and my first idea for proving it was assuming $\prod\limits_{n=1}^{\infty}(1+a_n)=S$, so $\sum\limits_{n=1}^{\infty}\ln(1+a_n)=\ln S$ converges but have no idea how to deal with this since $1+a_i$ is complex.

If this diverges in general, I was wondering if there are any necessary and sufficient conditions for the sum to converge such as $\sum\limits_n|a_n|<\infty$ or that the $a_n$ are positive real numbers.

Any help would be awesome,and thanks for taking a look!

$\endgroup$
  • $\begingroup$ I believe $\sum \ln(1+a_n) $ converging implies $\sum a_n$ converges, even for complex numbers, but I'm still looking for the proof (@twakhare it was, people keep giving diverging to 0 products). $\endgroup$ – user24142 Aug 11 '15 at 0:49
7
$\begingroup$

The infinite product $\prod_n (1+a_n)$ converges (to a nonzero value) if and only if $\sum_n \log(1+a_n)$ converges (for a branch of $\log$ that is analytic in a neighbourhood of $1$, with $\log(1) = 0$). This does not in general imply $\sum_n \dfrac{a_n}{1+a_n}$ converges. For example, take a positive sequence $b_k \to 0 + $ and let $c_k = 1/(1+b_k) - 1$. Thus $\log(1+c_k) = -\log(1+b_k)$. On the other hand, $$ \dfrac{c_k}{1+c_k} + \dfrac{b_k}{1+b_k} = \dfrac{-b_k^2}{1+ b_k} \ne 0$$ Consider a sequence $a_n$ that, for each $k$, repeats $b_k, c_k$ enough times for the partial sum of $a_n/(1+a_n)$ to change by $1$, before progressing to the next $k$. Then $\sum \log(1+a_n)$ converges to $0$ but $\sum a_n/(1+a_n)$ diverges.

$\endgroup$
  • 1
    $\begingroup$ Very nice. A simple sequence constructed using this idea is $a_n =\{\color{red}{\frac{1}{2},-\frac{1}{3},\frac{1}{2},-\frac{1}{3}}$, $\color{blue}{\frac{1}{3},-\frac14,\frac13,-\frac14,\frac13,-\frac14}$, $\color{green}{\frac14,-\frac15},\color{green}{\frac14,-\frac15},\color{green}{\frac14,-\frac15},\color{green}{\frac14,-\frac15},\ldots\}$. Then $\prod(1+a_n) = \color{red}{1}\cdot \color{blue}{1}\cdot \color{green}{1}\cdots$ while $\sum\frac{a_n}{1+a_n} = -\left(\color{red}{\frac{1}{3}} + \color{blue}{\frac{1}{4}} + \color{green}{\frac{1}{5}} + \ldots\right)$. $\endgroup$ – Winther Aug 11 '15 at 1:36
2
$\begingroup$

It's OK for absolute convergence. But not in general.

Take $\log(1+a_n) = (-1)^n/\sqrt{n}$ for $n \ge 1$, so $\sum \log(1+a_n) = \sum (-1)^n/\sqrt{n}$ converges (conditionally). But $$ \frac{a_n}{1+a_n} = \frac{(-1)^n}{\sqrt{n}} - \frac{1}{2n} +O\left(\frac{1}{n^{3/2}}\right) $$ and thus $\sum a_n/(1+a_n)$ diverges.

$\endgroup$

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.