16

I'm familiarizing myself with manifolds. I tried to show $[0,1]\times[0,1]$ is a manifold with a boundary. Can you please tell me if my proof is correct:

The definition for manifold with boundary:

A manifold with boundary $M$ is a second countable Hausdorff space so that for a $p \in M$ there is an open set $U \subseteq M$ so that there is a homeomorphism $\varphi$ to either (a) an open set $V$ in $H^n \setminus \partial H^n$ or to (b) an open set $V$ in $H^n$ and $\varphi (p) \in \partial H^n$ where $H^n$ is the closed upper half plane. It means $H^n = \{x \in \mathbb R^n : x_k \ge 0 ; 1 \le k \le n \}$.

One has to show $M = [0,1] \times [0,1]$ is second countable, Hausdorff and locally homeomorphic to $H^n$.

$M$ is second countable because it has subspace topology of $H^2$.

It is Hausdorff because it has subspace topology of $H^2$.

Locally Euclidean: For point $p \in M$ take set $U$ open in $H^2$ with $p \in U$. Inclusion map $i: U \to U \subseteq H^2$ is local homeomorphism with the property maps to open set and if $p \in \partial M$ then $i(p) \in \partial H^2$.

I am very grateful for your help.

goobie
  • 541
  • 2
    Well, it is not a 2-manifold, but a 2-manifold with boundary. – Berci Jan 23 '13 at 14:01
  • @Berci Okay. Does it imply my proof is wrong? – goobie Jan 23 '13 at 14:05
  • 2
    What is your definition of manifold with boundary ? The usual definition is this one. If $x$ is on the boundary, then you are showing that the boundary (NOT the square) is a manifold locally at $x$. – user10676 Jan 23 '13 at 14:38
  • http://math.stackexchange.com/questions/5906/is-triangle-a-manifold?rq=1 – Myshkin Jan 23 '13 at 14:48
  • @user10676 I use this definition but with additional requirement that it is second countable. As I understand for interior points of $n$-manifold I have to show that there is neighborhood homeomorphic to $\mathbb R^n$ and for boundary points there is neighborhood homeomorphic to $\mathbb R^{n-1}$. Is it false? – goobie Jan 23 '13 at 14:50
  • 1
    No. You have to show that there is a neighborhood homeomorphic to $\mathbb{R}^{n-r} \times \mathbb{R}_{\geq 0}^{r}$ for some $r$ (depending on $x$). – user10676 Jan 23 '13 at 15:09
  • @user10676 I posted corrected work in the answer. Is it now correct? Thank you for help. – goobie Jan 25 '13 at 20:09
  • I suggest you to take the following definition. A topological (Hausdorff, second countable) space $M$ is a manifold of dimension $n$ iff : for all $x \in M$ there exist a $U \subset M$ open containing $x$, an integer $r \in [0,n]$, a $V \subset \mathbb{R}^{r} \times \mathbb{R}_{\geq 0}^{n-r}$ open contaning $0$, and an homeomorphism $\phi : U \rightarrow V$ sending $x$ to $0$. – user10676 Jan 27 '13 at 12:56
  • @user10676 I do not see the benefit in using a different definition. – goobie Jan 28 '13 at 10:18
  • There is something I don't understand in your definition and your proof. 1) What is $H^n$ ? I guess that $H^2$ is $\mathbb{R} \times \mathbb{R}_{\geq 0}$, but what about $n>2$ (or $n=1$) ? 2) In your proof $U$ is supposed to be a neighborhood of $p$ in $M$, but you take $U$ open in $H^2$, so something is wrong. – user10676 Jan 28 '13 at 10:32

1 Answers1

6

The portion where you show that $M$ is locally Euclidean is not correct and here's why:

You have to show that for every point $p \in M$ there exists a neighborhood $U \subseteq M$ of $p$ and a homeomorphism from $U$ to an open subset of either $\mathbb R^2$ or $\mathbb H^2$.

Now here's what you wrote:

  • For point $p \in M$ take set $U$ open in $\mathbb H^2$ with $p \in U$

So your neighborhood $U$ is in $\mathbb H^2$ and not in $M$! While it is indeed true that $M \subseteq \mathbb H^2$ we cannot simply say that a neighborhood in $\mathbb H^2$ is a neighborhood in $M$. If $p$ is the point $p = (1, 1)$ then any neighborhood of $p$ in $\mathbb H^2$ will contain points that are not in $M$.

Edit:

To fix the proof remember that if $U \subseteq M$ contains part of the boundry of $M$ then the chart for $U$ will have to map that boundry to the boundry of $\mathbb H^2$. So for example if $p = (1, 1)$ then let $U = (.4, 1] \times (.4, 1]$ and for a chart try $\phi\colon U \to \mathbb H^2$ defined by $\phi(x, y) = (1 - x, 1 - y)$. I leave it to you to check that the image of $\phi$ is open in $\mathbb H^2$ and that $\phi$ is a homeomorphism onto this image. You'll also need to come up with charts covering the rest of $M$. I suggest you take

  • $U_2 = [0, .6) \times (.4, 1]$
  • $U_3 = (.4, 1] \times [0, .6)$
  • $U_4 = [0, .6) \times [0, .6)$

as the neighborhoods for those charts.

Edit #2:

Whoops! My edit above has a mistake as pointed out by goobie. Also pointed out by goobie is the fix: Instead of the chart $\phi$ that I suggested take $\phi\colon U \to \mathbb H^2$ defined by $\phi(z) = z^2$ (here I'm using complex numbers to denote points in the plane). Then you'll just need to do some translation and rotation to handle the other corners.

Jim
  • 30,682
  • Thank you very much for answer! I understand mistake. To correct it one can take intersection of $U \cap M$ where $U$ is open in $H^2$? – goobie Jan 29 '13 at 07:35
  • No, if $U$ is a neighborhood of $p$ in $\mathbb H^2$ then the intersection $U \cap M$ will indeed be a neighborhood of $p$ in $M$, but it will not be an open subset in $\mathbb H^2$ so the inclusion of $U \cap M$ into $\mathbb H^2$ does not give a homeomorphism of $U \cap M$ onto an open subset of $\mathbb H^2$. – Jim Jan 29 '13 at 07:55
  • Many thank you, I can do it now!! But: if you take $U_1 = (0.4, 1] \times (0.4, 1]$ and denote its corners by $ABCD$ counterclockwise and then apply $\pi (x,y) = (1-x, 1-y)$ the result is a set with corner $C$ at $(0,0)$ and corners $CDAB$ counterclockwise. It has closed border $BC$ and is not open in $\mathbb H^2$. – goobie Jan 29 '13 at 09:12
  • I tried with open balls: $B( (1,1) , 0.6) \cap M$. Then denote a point in $B( (1,1) , 0.6) \cap M$ with radius and angle relative to $(1,1)$ as $(r_{(1,1)}, \varphi_{(1,1)})$. Then define map into $H^2$ absolute coordinates as $(r_{(1,1)}, \varphi_{(1,1)}) \mapsto (r_{(1,1)}, 2(\varphi_{(1,1)} + \pi / 2))$. – goobie Jan 29 '13 at 09:14
  • How can I award bounty to you? If I click I get this: http://i.stack.imgur.com/0pnU6.png Do I have to wait until bounty time is ending? I will try again later. – goobie Jan 29 '13 at 09:19
  • The image of $U_1$ under that map is $[0, .6) \times [0, .6)$. This is the intersection of the open set $(-.6, .6) \times (-.6, .6)$ with $\mathbb H^2$ and therefore is open in $\mathbb H^2$. – Jim Jan 29 '13 at 16:33
  • But $\mathbb H^2 = { (x,y) : y \ge 0 }$ is the upper half plane. Intersecting $(-0.6, 0.6) \times (-0.6, 0.6)$ with $\mathbb H^2$ yields the set $(-0.6, 0.6) \times [0, 0.6)$. It is open but not equals $[0,0.6) \times [0, 0.6)$. – goobie Jan 29 '13 at 19:00
  • Whoops! You're right. I was taking $\mathbb H^2 = {(x, y) \ | \ x, y \geq 0}$. – Jim Jan 29 '13 at 19:44