# Why isn't a lemniscate a manifold?

I would like a formal, but not very deep in the theory, answer to this question.

Maybe I am even wrong at the understanding of what a lemniscate may be, so here is another question:

Is the image of the function $f:[-2\pi,2\pi]\rightarrow \mathbb{R^2}$ given by:

$f(t)=(1+\cos(t),\sin(t))I_{(0,2\pi]}(t)+ (-1+\cos(t),\sin(t))I_{(-2\pi , 0)}(t)$

a lemniscate? (where $I_{X}$ is the indicator function of the set $X$)

-
Lemniscates are never manifolds because they cross over themselves and locally around the cross point it is not homeomorphic to $\mathbb{R}$ – Sean Ballentine Feb 19 '13 at 21:59
Yes, but why can't I get the homeomorphism? I understand the intuition, but as I said, I want something more formal. – Aloizio Macedo Feb 19 '13 at 22:04

Suppose we try to chart near the cross point, so we have a homeomorphism $f:(-\epsilon,\epsilon) \to X \subset \mathbb{R}^2$ where $X$ is the small cross. Then if we remove the single point in the center of the cross in $X$ and the corresponding point in $(-\epsilon,\epsilon)$, we get 4 connected components in the image but only 2 in the pre-image. Since homeomorphisms must preserve the number of connected components, no such homeomorphism is possible.

-
Formal answer and satisfactorily introductory. Upvoted and accepted, thanks! – Aloizio Macedo Feb 19 '13 at 22:12

As said in the comments, the lemniscate crosses over itself and thus cannot be homeomorphic around this point to $\mathbb{R}$. One way to see this is to think of what happens when the point is removed: you get 4 connected components in the lemniscate but only two in $\mathbb{R}$, so any map between the two cannot be an homeomorphism.

-

Your set is the union of two touching circles. One can easily give an explizit homeomorphism with $\mathbb R^1$ (in fact by means of $f$) of a sufficiently small connected open neighbourhood of any point except the origon. But this is not possible (nor is it possible with any $\mathbb R^n$) for the origin:

Let $U$ be an neighbourhood of $(0,0)\in X\subseteq \mathbb R^2$, $V$ an open neighbouthood of $O\in\mathbb R^n$ and $h\colon U\to V$ a homeomorphism with $h(0,0)=O$. Wlog. $U=X\cap B_{\epsilon}(0,0)$ for some $\epsilon<1$. Then $U\setminus \{(0,0)\}$ has exactly four connected components, but the supposedly homeomorphic image $V\setminus \{O\}$ is connected (if $n>1$) or has two components (if $n=1$), contradiction.

-
Why do we say "an neighborhood," instead "a neighborhood?" I've seen other people say that. I am starting to guess it’s on purpose. @Eitzen – Student Feb 19 '13 at 22:22
@George It really should be "a neighborhood". – Ragib Zaman Feb 19 '13 at 22:36