That is indeed a misconception, but my additional hypothesis makes it right. But I want you to be precise. What do I add to $s_{2n}$ to get $s_{2n+1}$?
OK, DogAteMy. Using only the definition of the derivative, show me that $f(\mathbf x) = \|\mathbf x\|$ is differentiable at any $\mathbf a\ne \mathbf 0$ and find the derivative. (I.e., you will need to show that the appropriate error term goes to 0.) As you'll find out, this function is one of my favorites.
So, @Null, now you are armed to consider your original question and a quest for a counterexample. You know that you can't have the terms going to 0. What's the easiest series you can think of where the terms don't go to 0 but the odd partial sums converge?
@Alessandro: If $f\colon\Bbb R^n$ is continuous along all curves coming into $0$, is it continuous at $0$? (We know it's false if it's known to be continuous only on lines or parabolas or .... ) Along those lines, can you give me a function that's continuous along all curves $|y| = |x|^n$ ($n\in\Bbb N$) but not continuous at $0$?
Alessandro: You keep blaming me for wasting your time when you have work to do ... and yet ... you ask for it!
Ah, I see. Find an infinite sequence of points which go to $0$ but whose values don't go to $f(0)$. Make the curve visit these points one at a time, finishing at the point $0$.
Thinking of the curve as the image of $\phi:[0,1]\to\Bbb R^n$, we let $\phi(1/n)$ be $a_n$, $\phi(0)$ be $0$, and at all other points interpolate linearly.
@TedShifrin Sure. It maps limit points to limit points, no?
http://math.stackexchange.com/questions/2090955/integration-using-substitution-int-frac1-sqrt1-x2-arcsin3x-dx/2090961#2090961. It seems like my least-effort posts always get the most upvotes.
I need to show that $GL(2,\mathbb{Z})=\left\{ \begin{pmatrix} a&b \\ c&d \end{pmatrix} \mid a,b,c,d \in \mathbb{Z}, ad-bc = \pm 1\right\}$ is not nilpotent. I have seen this question but the answer given there is too advanced for where I am currently in my studies.
In order to show something is ...
I think I can construct one that works up to $|y|=c|x|^n$ for a fixed $n$, let's work with functions $\Bbb R^2\to\Bbb R$ because they're easy to visualize. If I want a function continuous in $0$ along every line I take the $2$ parabolas $y=x^2$ and $y=-x^2$ and define $f$ as $1$ if $(x,y)$ lies on or above the graph of the first or on or below the graph of the second and $0$ everywhere else
in the same way if I want that to work for the families up to $c|x|^n$ I use the graphs of $y=|x|^{n+1}$ and $y=-|x|^{n+1}$ instead of the $2$ parabolas
Now, the second center is $Z_2 = \{ x \in \operatorname{GL}(2,\mathbb{Z}) \mid \forall y \in \operatorname{GL}(2,\mathbb{Z}) : [x,y] \in Z(\operatorname{GL}(2,\mathbb{Z}) )\}$.
eh, that's a problem. Because I can't use a function which is smaller than $x^n$ for every $n$ in a neighbourhood of $0$ since that would have to be constant
I did it for a function $\Bbb R^2\to\Bbb R$ because it's nicer to visualize. You draw the graph of $g(x)=e^{-1/|x|}$ and $h(x)=-e^{-1/|x|}$ and define $f(x,y)$ as $1$ if $(x,y)$ lies on or above the graph of $g$, on or below the graph of $h$ or on the $x$-axis. $f(x,y)=0$ in all other cases @Akiva
@TedShifrin do you mean, that at the $2n+1$th term the sum is zero, but one after that it is -1? (or 1, and 0) but that wouldnt be convergent. thatswhy i ask