# $\sum a_n$ be convergent but not absolutely convergent, $\sum_{n=1}^{\infty} a_n=0$

Let $\sum a_n$ be convergent but not absolutely convergent, $\sum_{n=1}^{\infty} a_n=0$,$s_k$ denotes the partial sum then could anyone tell me which of the following is/are correct?

$1.$ $s_k=0$ for infinitely many $k$

$2$. $s_k>0$ and $<0$ for infnitely many $k$

3.$s_k>0$ for all $k$

4.$s_k>0$ for all but finitely many $k$

if we take $a_n=(-1)^n{1\over n}$ then $\sum a_n$ is convergent but not absolutely convergent,but I don't know $\sum_{n=1}^{\infty} a_n=0$? so I am puzzled could any one tell me how to proceed?

• experiment with $-sum+1-1/2+1/3-1/4+1/5\cdots$ and see which answer is likely. – Maesumi May 23 '13 at 15:53
• For 2) I would say Sk > 0 OR Sk < 0 How can it be greater and less at the same time? I would rule out number 1) since a Conditionally Conv Series can sum up to anything depending on the arrangement. Let's see what others have to say. – imranfat May 23 '13 at 15:54
• By the way, for $4.$ try to understand what does it exactly mean. And find any series for which it is not satisfied – Ilya May 23 '13 at 15:56
• For (3) and (4), just note that if $a_k$ has this property then $b_k=-a_k$ has the same property. – Thomas Andrews May 23 '13 at 15:57
• I have edited $4$, but I am confused as I am not getting an example for which $\sum a_n=0$ – miosaki May 23 '13 at 16:01

None of them are necessarily true.

We can easily compute a series from its partial sums, so let's specify the $s_k$.

Define  s_k=\left\{\begin{array}{} -\frac1k&\text{if $k$ is odd}\$4pt] -\frac1{k^2}&\text{if k is even} \end{array}\right.  Then a_1=-1 and for k\gt1,  a_k=\left\{\begin{array}{} \frac1{(k-1)^2}-\frac1k&\text{if k is odd}\\[4pt] \frac1{k-1}-\frac1{k^2}&\text{if k is even} \end{array}\right.  Show that this series is not absolutely convergent, its sum is 0, and it fails to satisfy any of the conditions. It is easy to see that the first statement is wrong. The idea for a counterexample is in 3. For the second statement use Riemann's rearrengement theorem, which (if you unterstood the proof) gives you the existence of such a series. For the third statement construct a_k such that s_k=(-1)^k \cdot \frac{1}{k}. The fourth look at the third. On the other hand all of the things can be true. For the first let (b_n)_{n\in \mathbb{N}} be an arbitrary null sequence. Define a_n via \[a_n= \begin{cases} b_k & 2k=n\\ -b_k & 2k+1=n \end{cases}$ We see that $\sum_{n=0}^\infty a_n = 0$ and furthermore $\sum_{n=1}^{2N+1} a_n=0$ for any $N\in \mathbb{N}$. If you chose a "slow" enough null sequence it won't be absolute convergent.

For the second to be true use the series constructed in the first part at 3.

For the third part make something that a subsequence of $s_k$ is $\frac{1}{k}$, and blow up the rest with $n$ terms with values $\pm \frac{1}{n}$ and let $n \to \infty$ while $k\to \infty$.

The fourth is solved by the third.

• i have already said about $s_k$ u missed – miosaki May 23 '13 at 16:09
• @miosaki sorry about that. added some stuff to my answer – Dominic Michaelis May 23 '13 at 16:26
• only (3) and (4) correct.right? – Math geek Jun 9 at 1:02