Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I was trying to glue two cylinders and then show that the resulting space is a manifold. Here is the first of my attempts:

The cylinders I denote $C_0 = S^1 \times [0,1]$ and $C_1 = S^1 \times [0,1]$. Let $T \subseteq R^3$ be the surface $$x = \cos \theta (R + r \cos \phi),$$ $$y = \sin \theta ( R + r \cos \phi) ,$$ $$ z = r \sin \phi$$ with $\theta, \phi \in [0,2\pi]$ and $S^1 \times [0,1]$ be $(\cos \theta , \sin \theta , \phi) $ with $\theta \in [0,2\pi]$, $\phi \in [0, \pi / 2]$.

Let $g: C_0\times \{0\} \cup C_1 \times \{1\} \to T$ be $$(\cos \theta, \sin \theta , \phi , n ) \to (\cos ( \pi n + 2\phi)(R+r\cos \theta), \sin (\pi n + 2 \phi)(R + r \cos \theta), r \sin \theta)$$

Then on the disjoint union $C_0 \times \{0\} \cup C_1 \times \{1\}$ I define the equivalence relation $$(\cos s_1, \sin s_1, x_1, n') \sim (\cos s_0, \sin s_0, x_0,n)$$ $$\iff$$ $$ ((s_1 = s_0) \land (x_1 = x_0) \land (n=n')) \lor $$ $$ ((s_1 = s_0) \land (x_1 = 0 \land x_0 = \pi / 2) \land (n=0 \land n'=1)) \lor $$ $$((s_1 = s_0) \land (x_1 = \pi / 2 \land x_0 = 0) \land (n=0 \land n'=1)). $$

Let $T$ be the surface of the torus in $\mathbb R^3$ and $f: X /\sim \to T$, $[x] \mapsto g(x)$.

Then $f$ is bijective continous and open.

Therefore $T$ and $X/\sim$ are homeomorphic and therefore $X/\sim$ is a manifold.

Does it work like this? And how to show $f$ is open and continous? I tried but failed. Thank you for correcting me.

share|cite|improve this question
I'm a little confused about your map $f$. Does $X/ \sim$ denote the two cylinders glued-together? Then I don't see how you can define $f$ by $[x]\mapsto x$, because $X/ \sim$ and $T$ have a priori nothing to do with each other (one is an abstract topological space, the other a subspace of $\Bbb{R}^3$). Maybe you should work with an explicit parametrization of $T$, e.g. via $\sin$ and $\cos$, and then define your map $f$. – Nils Matthes Jan 17 '13 at 14:21
@NilsMatthes Is it more clear? I edited my question. – goobie Jan 17 '13 at 14:44
The relation you define isn't an equivalence relation, since it's not reflexive. – joriki Jan 17 '13 at 15:08
@joriki Thank you, I corrected it. – goobie Jan 17 '13 at 15:17
Is it okay to also post my result in an answer? – goobie Jan 17 '13 at 16:50
up vote 4 down vote accepted

If you want to prove that the gluing together of two cylinders is a manifold, you don't need to actually write too many things explicitly in the form of formulas. Even though the question doesn't explicitly state this, I take it from what you tried to do that the gluing occurs along the two circles $S^1\times\{0\}$ and along the circles $S^1\times\{1\}$.

To show that the result is a manifold, you don't explicitly need to know it is a torus (it is useful to realise this, but you won't need it). So, what do you do? First you describe the gluing process as $T=C_0\coprod C_1/\sim$, where $\sim$ identifies all relevant circles. You can do this as you did, but the same result can be obtained by only using two coordinates: one for the height, one for the angle.

When is a topological space a manifold? When it is locally homeomorphic to $\mathbb{R}^n$, in this case with $n=2$. So, this is what we need to show and we proceed by using the quotient topology on $T$: a set is open in $T$ if its preimage under the projection $\pi: C_0\coprod C_1\to T$ is open, i.e. if it is the disjoint union of an open in $C_0$ and one in $C_1$.

We now choose a point $p$ in $T$ and notice that there are two options:

1) The point `comes from' an interior point in one of the $C_i$, say $C_0$. In that case, the point in $C_0$ has an open neighbourhood $U$ away from the boundary of $C_0$. Moreover, this neighbourhood is a homeomorphism to an open part of $\mathbb{R}^2$. This neighbourhood is an open set in $T$ as well, so gives a local homeomorphism around $p$.

2) The point $p$ comes from the identified boundary points of the cylinders. In this case things get a bit more difficult. We need to find open sets $U_0$, $U_1$ in $C_0$, resp. $C_1$ that correspond on the boundary. There exists homeomorphisms of $U_0$ to an open part of $\mathbb{R}\times\mathbb{R}_{\leq0}$ and of $U_1$ to an open part of $\mathbb{R}\times\mathbb{R}_{\geq0}$. These can be put together to form an open part of $\mathbb{R}^2$, homeomorphic to $\pi(U_0\coprod U_1)$.

Parts 1 and 2 together show that local homeomorphisms to $\mathbb{R}^2$ exist and that the glued together cylinders form a manifold. It is however used that cylinders are manifolds with boundary. If you don't know this yet, you could more or less do the same to prove this, gluing together two sides of $[0,1]^2$.

share|cite|improve this answer
Thank you!! I have managed to almost prove that the resulting space is a manifold. Can I also post my work in an answer? – goobie Jan 17 '13 at 16:40
But to show that the resulting space is a manifold one also must show that it is Hausdorff and second countable. (and locally Euclidean)? – goobie Jan 17 '13 at 17:00
Having a local homeomorphism to some $\mathbb{R}^n$ is the same as being locally Euclidean, so that's exactly what I just did. Knowing the topology, I don't think proving it is Hausdorff will be too much trouble either - it is very similar to constructing the homeomorphisms above. For secound countability, you can look at bases for the topologies on both cylinders - which are second countable since $[0,1]^2$ is. A disjoint union of two countable bases is still countable and I don't think gluing will drastically change that. – HSN Jan 17 '13 at 17:09

Here is my work (with explicit formulas of imbeddings):

Let $$ \left ( \begin{array}{c} \cos \theta ( R + r \cos \varphi) \\ \sin \theta ( R + r \cos \varphi ) \\ r \sin \varphi \end{array} \right )$$ where $\theta, \varphi \in [0, 2 \pi]$ be a parametrization of the torus $T \subseteq \mathbb R^3$.

Let $(\cos a, \sin a , b ) $ where $a \in [0,2\pi], b \in [0,\pi / 2]$ be a parametrization of the cylinder $S^1 \times [0,1]$.

On $X = C_0 \sqcup C_1 = \{ (\cos a , \sin a, b , n ) \mid a \in [0,2\pi], b \in [0,\pi / 2], n \in \{0,1\}\}$ define

$$ (\cos a, \sin a, b, n) \sim (\cos a' , \sin a', b', n') \iff ((a = a') \land (b=b') \land (n=n')) \lor ((a =a') \land (b=\pi/2) \land b'=0 \land (n=0) \land (n'=1)) \lor ((a =a') \land (b=0) \land b'=\pi/2 \land (n=0) \land (n'=1))$$

Let $f: C_0 \sqcup C_1 \to T$ be the map $$ (\cos a , \sin a, b, n) \mapsto \left ( \begin{array}{c} \cos ( \pi n + 2 b)(R + r \cos a) \\ \sin (\pi n + 2 b)(R + r \cos a) \\ r \sin a \end{array} \right )$$ and define $g: X/\sim \to T$ as $[x] \mapsto f(x)$.

Here $g$ is (1) welldefined, (2) injective (3) surjective, (4) continous and (5) open:

(1) welldefined: Assume $[x]=[y]$ and without loss of generality, $x \in C_0$. For $x \neq y$ this happens in two cases: either $$ x = (\cos a, \sin a, 0,0) , y = (\cos a, \sin a, \pi / 2, 1)$$ or $$ x = (\cos a, \sin a, \pi / 2,0) , y = (\cos a, \sin a, 0, 1)$$ In the first case, $$ g([x])=f(x) = \left ( \begin{array}{c} \cos ( \pi \cdot 0 + 2 \cdot 0)(R + r \cos a) \\ \sin (\pi \cdot 0 + 2 \cdot 0)(R + r \cos a) \\ r \sin a \end{array} \right ) = \left ( \begin{array}{c} \cos ( \pi \cdot 1 + 2 \cdot \pi / 2)(R + r \cos a) \\ \sin (\pi \cdot 1 + 2 \cdot \pi / 2)(R + r \cos a) \\ r \sin a \end{array} \right ) = g([y])$$ Similarly in the second case. Hence $g$ is welldefined.

(2) $g$ is injective: Assume $g([x]) = g([y])$ where $x = (\cos \theta_x, \sin \theta_x, \varphi_x,n_x)$ and $y = (\cos \theta_y, \sin \theta_y, \varphi_y,n_y)$. Then $r \sin \theta_x = r \sin \theta_y$ hence $\theta_x = \theta_y$. From this it follows that $\cos(\pi n_x + 2 \varphi_x ) = \cos(\pi n_y + 2 \varphi_y ) $ and $\sin(\pi n_x + 2 \varphi_x ) = \sin(\pi n_y + 2 \varphi_y ) $. This implies $\pi n_y + 2 \varphi_y = \pi n_x + 2 \varphi_x $. If $n_x = n_y$ it follows that $\varphi_x = \varphi_y$. If $n_x \neq n_y$, without loss of generality assume $n_x = 0$ and $n_y = 1$. Then $\pi + 2 \varphi_y = 2 \varphi_x$. Since $\varphi_x, \varphi_y \in [0, \pi / 2]$ this implies that $\varphi_y = 0$ and $\varphi_x = \pi / 2$. But then also $[x]=[y]$.

(3) $g$ is surjective: Let $$y = \left ( \begin{array}{c} \cos ( \pi n + 2 b)(R + r \cos a) \\ \sin (\pi n + 2 b)(R + r \cos a) \\ r \sin a \end{array} \right ) \in T$$ then for $x = (\cos a , \sin a, b, n)$, $g([x]) = y$.

(4) $g$ is continous: Let $q : X \to X/\sim$ be the quotient map. Then $q$ is open by the definition of the quotient topology. Hence $g = f \circ q^{-1}$ is continous as the concatenation of two continuous functions.

(5) $g$ is open: $g = f \circ q^{-1}$ and $q$ is open and continous hence $g$ is open $\iff$ $g \circ q = f$ is.

Let $U$ in $X$ be open. Then $U = U_0 \sqcup U_1$ where $U_0, U_1 $ are open and $U_0 \subseteq S^1 \times [0,1] \times \{0\}$ and $U_1 \subseteq S^1 \times [0,1] \times \{1\}$. $f$ restricted to $U_0, U_1$ respectively is the identity map hence open. Then $f(U) = f(U_0) \cup f(U_1) $ is open.

Therefore $g$ is a homeomorphism.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.